Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2026 Jul 1;18(4):e70179. doi: 10.1111/aphw.70179

Examining the clustering of lifestyle factors and affect in daily life: An idiographic approach

Austen R Anderson 1,2,✉, Lindsey Ostermiller 2, Isabel Rice 2, Casandra O'Rourke 1,2
PMCID: PMC13321139  PMID: 42383508

Abstract

There has been an increase in interest in the health and well‐being benefits of lifestyle factors such as physical activity, diet, sleep, and social interaction. Previous research has highlighted how lifestyle factors, both healthy and unhealthy, tend to covary or cluster together. Very little research has examined the degree to which the clustering among lifestyle factors and emotional health is unique to individuals. As such, this project examined the idiographic associations among lifestyle behaviors and emotional experiences in daily life, which could set the stage for individualized lifestyle interventions. Seventy‐nine adults (M age  = 43.37, SD age  = 17.26) in the United States participated in a 70‐day daily diary study, where they reported on daily lifestyle engagement and affect each evening. Network models examined the associations among these variables at the between‐person level and within‐person level. Further, person‐specific idiographic network models were analyzed. Results indicate that there were significant associations in the whole sample at multiple levels (between‐person and within‐person). Further, idiographic analyses revealed person‐specific connections among lifestyle factors and affect. In conclusion, lifestyle factors and affect cluster together in daily life, and the clustering differs across individuals. Future research should explore the feasibility of intensive measurement‐informed lifestyle interventions.

Keywords: daily diary method, health behavior, individual differences, lifestyle, mental health

INTRODUCTION

In recent decades, there has been an increase in interest in healthy lifestyle factors and a growing focus on health and wellness in daily life (Firth et al., 2020; Walsh, 2011). Adherence to healthy lifestyle factors including limiting alcohol use, sufficient engagement in physical activity, maintaining a healthy diet, avoiding tobacco use, and maintaining good sleep quality has been associated with lower mortality rates and increased life expectancy (Li et al., 2020, 2023; Zhang et al., 2020). Further, a lifestyle medicine perspective has also been applied to emotional health, showing that lifestyle factors are important for the initiation/prevention and maintenance of psychological distress (Firth et al., 2020). These lifestyle factors are expected to operate through various psychological and physiological mechanisms to promote health (El‐Malahi et al., 2024). The potential impact of lifestyle on health is heightened with studies showing that lifestyle factors cluster together cross‐sectionally (Meader et al., 2016; Nudelman et al., 2019) and over long periods of time (Anderson et al., 2022), with some early evidence showing clustering occurring at the daily level (Anderson & Fowers, 2020). Nonetheless, little is known about the ways in which a broad array of lifestyle factors cluster together in daily life and whether there might be important individual differences in the ways in which lifestyle factors cluster together. As such, the present study explored the clustering of lifestyle factors using daily diary data over a period of 70 days, including an investigation of person‐specific clustering using network models.

Between‐person clustering of lifestyle factors

The clustering of lifestyle factors has been demonstrated at multiple levels of analysis. Most commonly, this clustering is explored in cross‐sectional survey research (Meader et al., 2016; Nudelman et al., 2019). This research indicates that, for example, individuals who eat better tend to exercise more whereas people who tend to drink more alcohol tend to use more tobacco. These cross‐sectional findings highlight how lifestyle behaviors tend to cluster together in a population, which could be useful to help identify populations with the greatest needs due to overall poor lifestyle profiles. However, these approaches do not allow for any kind of causal claims or an examination of how changes in lifestyle factors may be associated with changes in other factors or mental health over time.

Conceptualizing how and why lifestyle factors cluster together can be carried out from various perspectives. At the between‐person level, healthy and unhealthy behaviors may co‐occur because of the impacts of stable social determinants of health (e.g. income, education, region, and sex). Further, stable individual differences in personality, resilience, or other psychological resources might impact a person's engagement in healthy or unhealthy behaviors. Although this between‐person perspective is helpful, especially from a public health perspective, it fails to examine fluctuations in these factors in daily life and their clustering within person.

Within‐person clustering of lifestyle factors

Other research has examined how variations in these variables over long periods of time are associated with each other (Chevance et al., 2020). For example, an increase in a person's physical activity from one time point to another might be associated with concomitant increases in social interaction or decreases in smoking (Anderson et al., 2022). This indicates that some combination of circumstances (moving to a healthier environment) or behavior change (choices around engagement in health‐related behaviors) may link changes in one behavior to another.

There is reason to believe that behavior and behavior change processes are best conceptualized using finer‐grained time points (van Os et al., 2017). Supporting this presumption, research has shown that fluctuations in lifestyle engagement co‐occur at the daily level, resulting in what are called within‐person contemporaneous associations where fluctuations in lifestyle factors on a given day are associated with fluctuations in other lifestyles and emotional health on that same day. As an example, spending more time in nature than one's average on a given day has been associated with greater physical activity and social interaction on that day (Anderson & Fowers, 2020). More recently, research has been examining the clustering of variables (e.g. mental health symptoms) using within‐person temporal associations, where fluctuations in a given variable at one time point are associated with fluctuations in another variable at the next measurement time point (Levinson et al., 2020).

A key reason to examine associations at both between‐ and within‐person levels is that the associations between the same set of variables can differ by strength and even direction, which is a form of the Simpson's paradox (Kievit et al., 2013). As an example, people who exercise more on average tend to have a lower risk of heart attacks (between‐person), but on days in which someone exercises more than their average, they have a greater risk of heart attack on that day (within‐person) (Mittleman et al., 1993). To understand the association between physical activity and heart attack risk, it would not be helpful to rely solely on one level of analysis and say that they are negatively (between‐person) or positively (within‐person) associated—too much information is lost. Thus, to fully understand lifestyle factors and how they relate both to each other and to emotional health, it can be helpful to examine these associations at different levels and using varying time scales as they each provide unique sources of information.

Conceptually, clustering at the within‐person level may be driven by various processes, including simply because pairs or groups of behaviors occur together, such as physical activity, which might occur as a walk around one's tree‐lined neighborhood (nature engagement), with a close other (social interaction; Hartig et al., 2014). Another perspective highlights that engaging in healthy activities may put someone in contact with health‐conscious peers, influencing other lifestyle choices (Latkin & Knowlton, 2015). The broaden and build theory argues that over time, individuals who experience positive emotions tend to have broader thought and behavior repertoires, along with more personal resources, which creates more positive emotions and a positive upward spiral of positive behaviors and emotions (Fredrickson, 2004). Positive behaviors (e.g. physical activity) on a given day might promote positive affect, which broadens thought and behavioral repertoires, supporting an individual in making more self‐regulated decisions about other lifestyle behaviors (Fredrickson, 2013). In fact, some evidence indicates that positive emotion alone can non‐consciously activate goals related to health behaviors (Cameron et al., 2018), reinforcing the upward spiral. These processes might be complemented by psychological momentum, which occurs when success at a goal‐oriented task breeds self‐confidence and competence, heightening expectations, increasing effort, and promoting a greater likelihood of success at subsequent goal‐oriented tasks. Completing a relevant task (“getting to bed on time”) might lead one to have greater motivation and feel more confident in their ability to complete other subsequent related tasks (“connect with friends”; Iso‐Ahola & Dotson, 2016). Alternatively, negative emotions are associated with avoidant, habitual, and restricted behaviors (Herz et al., 2020), potentially contributing to downward spirals of negative emotions, narrow thought and behavior repertoires, and poor behaviors (sedentary behavior and poor diet). Interestingly, counterintuitive processes underlying lifestyle factor clustering may also be at play. Positive mood can increase risk‐taking (Saka & Yildirim, 2024) and compensatory health beliefs that engaging in one unhealthy behavior (regular drinking) can be compensated through engagement in another healthy behavior (healthy eating; Knäuper et al., 2004), which may link some of these behaviors together in daily life in counterintuitive ways.

Person‐specific clustering of lifestyle factors

Although there is clear evidence that lifestyle factors cluster together at the population level, and growing evidence that lifestyle factors fluctuate in daily life, a generally unanswered question is whether daily fluctuations and clustering of lifestyle factors differ by person. A great variety of personal factors, ranging from culture (James, 2004), developmental processes (Dennaoui et al., 2024), values (Brunsø et al., 2004), past experiences with lifestyle factors (Blanchard et al., 2023), socioeconomic factors (Giles‐Corti & Donovan, 2002; Olano et al., 2015), and personality (Ferdosi et al., 2020), can all impact lifestyle engagement and potentially how lifestyle factors cluster together. The question about person‐specific lifestyle clustering is aligned with previous research that highlights the idiographic nature of mental health symptoms, with unique patterns of clustering occurring over time for each person (Epskamp, van Borkulo, et al., 2018). So, although clustering of lifestyle factors in daily life may vary significantly for each person, this remains an unanswered question. To demonstrate such clustering requires specific methodological approaches.

Intensive longitudinal designs, which collect a high number of data points per person, can make it possible to examine idiographic network models of lifestyle factor clustering along with between‐person and within‐person networks. As an example of idiographic within‐person clustering, greater than average physical activity may be positively associated with next‐night sleep quality for one individual but not associated or even negatively associated with next night sleep quality for another individual or the sample on average. Such an idiographic approach to lifestyle clustering may have important implications for applied settings as recent work has evaluated the use of idiographic models in clinical settings both in service of tailoring treatment and delivering personalized feedback to participants (Harris et al., 2025; Levinson et al., 2023; Scholten et al., 2022). For example, Fisher et al. (2019) used intensive data collection before treatment to personalize a modular treatment for mood and anxiety. In other studies, idiographic networks of symptoms were deemed to increase insight and contribute to case conceptualization (Bootsma et al., 2022; Frumkin et al., 2021). The emerging research suggests that idiographic models can support intervention personalization and potential effectiveness. The present study lays some of the preliminary groundwork for implementing idiographic models within a lifestyle‐oriented approach to improving mental health.

Network models can be used to delineate complex associations among variables across levels of analysis (Borsboom et al., 2021). They can be used on cross‐sectional data to demonstrate clustering at the between‐person level, and, when used with intensive longitudinal data, they can assess clustering at both the between‐ and within‐person level in daily life—where daily fluctuations of lifestyle behaviors around person averages of those lifestyle behaviors can cluster together in a sample (Chevance et al., 2020; Epskamp, Waldorp, et al., 2018). The most recent applications of network analysis can, with large numbers of measurement points per person, extend the examination of clustering to person‐specific within‐person processes, in conjunction with between‐person and sample‐wide within‐person associations. This approach has opened new avenues for more fully grasping the complexity of lifestyle and emotional health clustering in daily life.

Present study

The central aim of the current study is to more extensively map the clustering of lifestyle factors and emotional health by identifying between‐person clustering, within‐person contemporaneous and temporal clustering, and at the idiographic level examine individual variation in contemporaneous within‐person clustering. This aim required the novel integration of an intensive longitudinal design examining lifestyle factors and emotional health in daily life in conjunction with multilevel network analyses. As such, the study demonstrates the utility of network analysis for high‐resolution lifestyle data and idiographic modeling. To ensure a robust understanding of the novel idiographic associations, two approaches to the analyses were included, both with strengths and weaknesses, which could have important implications for future research on idiographic networks in clinical settings. These approaches are discussed more thoroughly in the analysis section. As demonstrated by the Simpson's paradox, examining lifestyle factor clustering at the between‐person level demonstrates aggregate level clustering and could provide important context for understanding within‐person clustering. Further, the sample‐wide within‐person associations will provide a point of comparison and contrast for the idiographic models and will also be used to directly inform the person‐specific networks in one of the two analytical approaches. Following previous research (Anderson & Fowers, 2020), our first hypothesis was that healthy lifestyle behaviors (e.g. fruit and vegetable intake, physical activity, social interaction, nature engagement, relaxation, and time spent in hobbies) would correlate positively with each other and have a positive effect in our sample at the between and within‐person levels, while correlating negatively with negative affect and sleep distress. The roles of other lifestyle factors (alcohol use, caffeine use, TV and movie watching, napping, and sleep duration) were viewed more exploratively. The second hypothesis is that there would be heterogeneity in the within‐person contemporaneous network structures across participants, demonstrating idiographic variation in lifestyle clustering in daily life.

METHODS

Participants

Participants were recruited from the online participant platform www.prolific.ac, which connects researchers to potential participants (Palan & Schitter, 2018). Users of the online platform receive compensation for their participation and are drawn from a convenience sampling pool of more than 38,000 users residing in the United States. To obtain a sample of diverse ages, recruitment was split among three age brackets 18–30, 31–55, and 56–80. One‐hundred twenty‐three adults consented to participate in the study and completed the baseline survey. After excluding those who did not complete more than one of the daily surveys, 98 participants were left. Due to the need for a greater number of days' worth of data per person, especially for the person‐specific models (Epskamp, Waldorp, et al., 2018a), only those with 10 or more responses were included in the final sample (n = 79). See Table 1 for full participant characteristics.

TABLE 1.

Participant characteristics.

Variable Percentage (N = 79)
Age
18–30 32.9% (n = 26)
31–55 29.1% (n = 23)
56–80 36.7% (n = 29)
Missing 1.2% (n = 1)
Gender
Female 45.6% (n = 36)
Male 53.2% (n = 42)
Prefer not to say 1.3% (n = 1)
Race
White 72.2% (n = 57)
Black and/or African American 6.3% (n = 5)
Native American/Alaskan Native 1.3% (n = 1)
Asian Indian 2.5% (n = 2)
Chinese 1.3% (n = 1)
Filipino 1.3% (n = 1)
Multi‐racial 12.7% (n = 10)
Other 1.3% (n = 1)
Missing 1.3% (n = 1)
Employment status
Nonworking student 3.8% (n = 3)
Working student 3.8% (n = 3)
Nonworking, not looking for work 7.6% (n = 6)
Nonworking, retired 16.5% (n = 13)
Nonworking, currently looking for work 6.3% (n = 5)
Nonworking, at home caretaker 3.8% (n = 3)
Full‐time worker 41.8% (n = 33)
Part‐time worker 10.1% (n = 8)
More than one part‐time position 5.1% (n = 4)
Missing 1.3% (n = 1)
Education status
<High school degree 1.3% (n = 1)
High school degree 8.9% (n = 7)
Some college, no college degree 25.3% (n = 20)
2‐year or vocational degree 8.9% (n = 7)
4‐year degree 36.7% (n = 29)
Graduate degree 19.0% (n = 15)
Open in a new tab

Procedure

Participants reviewed a description of the study on Prolific, and interested participants were directed to an online survey delivered by Qualtrics that contained a digital consent form. Participants provided consent and then completed a baseline survey with demographic items. After completing the baseline survey, participants were sent a daily email through Prolific and were instructed to take an online survey between 8:30 and 11:59 p.m. (based on their local time) each night, for the next 70 days. The goal to recruit 100 participants for a 70‐day study was in line with Epskamp, van Borkulo, et al. (2018) simulation studies, showing adequate performance in detecting true networks, especially with the mlVAR approach. To maximize data retention, responses that were completed between 7:00 p.m. and 2:00 a.m. were included in the analysis. The procedure resulted in 4018 days' worth of data, for an average of about 51 days per participant. At the end of 70 days, participants were paid $15–$128 according to their level of completion. The study was reviewed by the Institutional Review Board at the University of Southern Mississippi.

Measures

Table 2 provides descriptive statistics and correlations for the measures included in the study.

TABLE 2.

Descriptive statistics for daily study variables.

Mean (SD) ICC 1 2 3 4 5 6 7 8 9 10 11 12 13 14
1. Physical activity a 3.63 (5.02) .42 1.00 .49 .09 .11 −.13 .24 −.11 .19 .47 −.04 .01 −.23 −.04 .33
2. Fruits & vegetables b 2.99 (2.35) .47 .11 1.00 .21 .09 −.07 .10 −.03 .09 .37 .01 .08 −.22 −.22 .38
3. Alcohol b 0.41 (1.13) .23 .05 .00 1.00 .15 .05 .06 .03 −.08 .11 −.01 .03 .02 .17 .13
4. Caffeine b 1.81 (1.64) .45 .08 .00 .00 1.00 .20 .13 .05 .14 .12 .06 −.16 .10 .03 −.05
5. TV and movies c 1.72 (1.70) .41 −.10 .01 .04 .01 1.00 −.09 .49 .05 −.04 .22 .13 .11 −.03 −.08
6. Hobbies c 1.10 (1.27) .40 .09 .11 .03 .04 .03 1.00 .21 .26 .13 −.16 .25 −.27 −.08 .13
7. Relaxing c 2.02 (1.65) .44 −.07 .03 .05 .00 .32 .19 1.00 .07 −.05 .21 .05 .11 .08 −.07
8. Social interaction c 2.15 (1.95) .42 .14 .05 .11 .06 −.10 .08 .01 1.00 .11 −.21 .19 −.27 −.24 .28
9. Nature engagement c 0.58 (0.90) .35 .31 .09 .07 .00 −.01 .08 .03 .18 1.00 −.06 .00 −.16 −.11 .26
10. Nap duration 0.27 (0.57) .33 −.08 −.02 −.01 −.05 .08 −.02 .09 −.06 −.02 1.00 −.28 .11 .05 −.17
11. Sleep hours 6.96 (1.66) .48 .03 .03 .00 −.05 .01 .03 .07 .01 .03 −.14 1.00 −.38 −.21 .19
12. Sleep distress 2.83 (2.54) .40 −.05 −.06 −.02 −.01 .00 −.05 −.04 −.06 −.04 .04 −.36 1.00 .35 −.47
13. Negative affect 4.20 (4.27) .40 −.08 −.10 −.03 −.01 −.02 −.11 −.11 −.17 −.11 .05 −.10 .20 1.00 −.56
14. Positive affect 13.90 (5.78) .52 .13 .15 .07 .05 −.03 .16 .12 .29 .16 −.07 .12 −.20 −.50 1.00
Open in a new tab

Note: Values above the diagonal represent between‐person bivariate correlations. Values below the diagonal represent within‐person bivariate correlations.

a

Metabolic units.

b

Servings.

c

Hours in that activity.

Sleep

Sleep quality, sleep quantity, and time spent napping were all assessed. Sleep quality was assessed using the Brief Pittsburg Sleep Quality Index (B‐PSQI; Sancho‐Domingo et al., 2021), which was modified to ask about past‐night sleep rather than the previous 30 days. The B‐PSQI is a six‐item measure that asks participants to report their bedtime and waketime, how long it took to fall asleep each night, predicted total sleep hours, how often the participant had trouble sleeping due to waking up in the middle of the night or early morning, and overall sleep quality. Sleep duration was included as a separate variable, so the other B‐PSQI items were summed, resulting in a score between 0 and 12, with higher scores representing greater sleep disturbances. An assessment of multilevel reliability indicated adequate reliability at the within‐person (ω w  = .66) and between‐person (ω b  = .78) levels. Participants were also asked to report on nap duration each day (none, 15 min, 30 min, 45 min, 1 h, and 2+ h).

Physical activity

Physical activity was measured using the Godin Leisure‐Time exercise questionnaire, which was modified for daily use (Flueckiger et al., 2017; Godin & Shephard, 1985). Participants reported on how much time they spent in mild, moderate, and vigorous physical activity. To create a total daily physical activity value, minutes of exercise were multiplied by a metabolic unit (3 for mild exercise, 5 for moderate exercise, and 9 for strenuous exercise) and summed to create one total daily physical activity value. The questionnaire has demonstrated adequate reliability and validity (Jacobs et al., 1993), and in this study, the intraclass correlation coefficient was .52, indicating that about half of the variation in the score was due to differences between people and half was due to day‐to‐day variation within people.

Diet

Diet was measured with two items asking participants to report how many servings of fruit and vegetables they consumed on average per day. Responses ranged from zero servings per day to more than six servings per day and summed to create a diet variable. Graphical examples of serving sizes for both fruits and vegetables were provided. For this two‐item composite, the between person reliability was adequate (ω b  = .77), whereas the within person reliability was lower (ω w  = .37).

Substance use

Alcohol use was measured by asking participants to report how many drinks of alcohol they consumed in the past 24 h. Participants were also asked to report how many caffeinated drinks they consumed each day with responses ranging from 0 to 10+.

Time usage

Six other variables related to participants' use of time were measured: enjoying nature, watching TV/Movies, socializing with close others, relaxing, and engaging in hobbies. Participants were asked how much time they spent engaging in each lifestyle behavior in 5‐min increments from 0:00 to 8:00+.

Affect

Positive and negative affect were measured with the Scale of Positive and Negative Experiences (SPANE; Diener et al., 2010). This 12‐item measure assesses emotional experiences and was modified to ask about daily emotional experiences. Responses to the items are on a 5‐point continuum, ranging from very rarely or never to very often or always. The measure is reliable and valid, demonstrating stronger associations with depression (and well‐being) than the widely used Positive and Negative Affect Schedule (Jovanović, 2015). Following Lai's (2021) approach, positive affect demonstrated acceptable reliability at the within‐person (ω w  = .89) and between‐person (ω b  = .97) levels, and negative affect also demonstrated acceptable reliability at the within‐person (ω w  = .82) and between‐person (ω b  = .89) levels.

Analyses

A network modeling approach was utilized to assess clustering of lifestyle factors and emotional health across multiple levels. Network models result in partial correlation networks, where each potential association represents the correlation between two variables, controlling for the influence of all other variables in the network. In particular, the graphical vector auto‐regression (GVAR) approach was implemented to account for the nested nature of intensive longitudinal designs and temporal ordering of effects. This modeling approach can provide sample‐wide associations among the variables at the between‐person level (e.g. people who spend more time in nature tend to exercise more), contemporaneously at the within‐person level (e.g. on days when people spend more time in nature than their average, they tend to exercise more than their average), and temporally at the within‐person level (e.g. days in which a person spends more time in nature than their average are associated with next‐day decreases in time spent in nature). The mlVAR package in the R statistical software was used to assess two sets of GVAR models (Epskamp et al., 2024). The first set of models included only the lifestyle factors, whereas the second set of models added positive and negative affect. Single‐day lags were implemented for the temporal effects. A GVAR approach that accounts for temporal effects also likely produces less biased contemporaneous effects. The models are graphically displayed using networks of nodes (variables) and edges (partial correlations). To handle the sleep variables, which represent the sleep from “last night,” the contemporaneous associations were set up so that last night's sleep would be associated with “today's” lifestyle factors. And then, the temporal associations were set up so that lifestyle factors “today” would predict “tonight's” sleep.

1 Most Gaussian‐based network analyses (including GVAR models) assume normality in the variables. It has been recommended to transform variables for gaussian networks using a nonparanormal method for estimating a semiparametric Gaussian copula models (Liu et al., 2009). The benefit of this approach, unlike traditional transformations (log, or square root), is that the conditional independence structure is preserved so the interpretation of the network is essentially the same. This transformation was carried out on the person averages of the variables, resulting in data that better approximated normality and can be readily interpreted.

A sensitivity model was run controlling for weekend‐weekday status, resulting in very little change to the findings. A second sensitivity model was run on the full sample, including those who completed less than 10 days of daily surveys. The sample‐wide within‐person models, which have greater precision due to the greater quantity of data, were very minimally impacted by the inclusion of the whole sample. The average size of the absolute difference in these partial correlations was .003 (SD = 0.003). The between‐person model's structure also generally remained the same, but as the between‐person sample size had a greater proportional increase going from 79 to 98, the average absolute difference in partial correlations was larger at about .04 (SD = 0.03). Overall, due to the limited differences and to avoid bias in the estimates (Epskamp, Waldorp, et al., 2018b), we used the sample of people who completed 10+ daily surveys, and the results for the full sample are presented in supplementary analyses.

To examine individualized contemporaneous network models, two separate approaches were used. The first approach again utilized the mlVAR package. This approach borrows information from the sample‐wide contemporaneous models, which is then updated with the participant's data to create an idiographic network of person‐specific associations. The second approach relied on the graphicalVAR package, which relies exclusively on each individual participant's' data, without reference to the wider sample. Overall, graphicalVAR models, which do not borrow information from other participants, are expected to perform less well in estimating within‐person associations when there is a limited number of observations per participant (Epskamp, Waldorp, et al., 2018a) but may provide insight into which associations are strongest for that participant.

RESULTS

Whole sample networks

The GVAR model containing only lifestyle factors revealed associations at the between, contemporaneous within‐person level, and temporal within‐person level (see Figure 1). Some of the associations are reviewed briefly below, but the full results are available in the figures and the Supporting Information. In line with our hypotheses, at the between‐person level, some of the positive lifestyle behaviors were correlated, including physical activity with fruit and vegetable intake (r p = .45) and with nature engagement (r p = .25). In turn, fruit and vegetable intake was associated with nature engagement (r p = .30). Sleep duration and relaxation seemed to have widespread associations with other variables (Figure 1, between‐person relations). At the contemporaneous within‐person level, there were significant partial correlations ranging from .04 to .42 in absolute value. Among those are associations that were found in a previous daily diary study (Anderson & Fowers, 2020) with time in nature being associated with physical activity (r p = .34, p < .001) and social engagement (r p = .08, p = .01), with physical activity being associated with fruit and vegetable intake (r p = .04, p = .01), and with alcohol use being associated with social engagement (r p = .07, p = .01). Because this study built on previous work by including an expanded set of lifestyle factors, other significant associations were apparent as well (Figure 1, within‐person contemporaneous relations). With the temporal associations, there were significant autocorrelations for all lifestyle factors except sleep duration, indicating that, for example, a higher than average amount of time spent on hobbies on 1 day was associated with a higher likelihood of time spent on hobbies on the next day (Figure 1, temporal [lag−1] relations). There were fewer temporal associations overall, however.

FIGURE 1.

FIGURE 1

Open in a new tab

Lifestyle factor clustering in the whole sample. Solid lines represent positive associations, whereas dashed lines represent negative associations. The thickness of the lines represents the size of the partial correlations. Only significant associations are shown. The models with color‐coded edges are also available as supporting information.

The GVAR model that added positive and negative affect revealed additional associations at each level of analysis. All findings and network figures are available in the Supporting Information. Additionally, the sample‐wide within‐person contemporaneous associations are seen in Figure 2, and the significant associations with affect are briefly reviewed. Daily positive affect was associated with social engagement (r p = .19, p < .001), nature engagement (r p = .07, p < .001), fruit and vegetable intake (r p = .06, p = .001), sleep duration (r p = .05, p = .01), alcohol intake (r p = .05, p = .002), time spent in hobbies (r p = .04, p = .02), and time spent relaxing (r p = .04, p = .03). Daily negative affect was associated with sleep distress (r p = .11, p < .001), time spent in hobbies (r p = −.05, p = .003), and time spent watching TV/movies (r p = −.04, p = .04). Negative and positive affect had a partial correlation of −0.40.

FIGURE 2.

FIGURE 2

Open in a new tab

The clustering of lifestyle and affect in the whole sample. Solid lines represent positive associations, whereas dashed lines represent negative associations. Only significant associations are shown. The thickness of the lines represents the size of the partial correlations.

Idiographic networks

Using the mlvar approach, individualized contemporaneous within‐person associations were modeled, with each individual model drawing information from the sample‐wide within‐person estimates. Sample models are displayed in the top row of Figure 3, and each individual model for the whole sample is available as Supporting Information. To aid in interpretability, the network models were constrained to only show partial correlations greater than .10. There is wide variation in how these variables cluster together across participants. Note, for example, the more central role of physical activity for Participant 65 whereas physical activity was only associated with nature engagement for Participant 28. Participant 29 also has a sparser network, with fewer edges over the .10 threshold, whereas Participants 20 and 65 both had denser networks. Figure 4 also displays cross‐participant differences in strength centrality, which is the sum of the weights of all connections (edges) linked to a given node.

FIGURE 3.

FIGURE 3

Open in a new tab

Person‐specific/idiographic within‐person contemporaneous network models of lifestyle and affect. Models from the two‐step multilevel VAR approach are on the top row, and individual LASSO models are on the bottom row. Solid lines represent positive associations, and dashed lines represent negative associations. The thickness of the lines represents the size of the partial correlations.

FIGURE 4.

FIGURE 4

Open in a new tab

Person‐specific/idiographic within‐person contemporaneous strength centrality indices. Standardized Z‐score values of strength centrality are displayed for each participant. Higher values represent a greater sum of absolute value partial correlations that variable has relative to other variables.

Using the graphicalVAR approach, individual models were estimated for each participant, with sample models provided for the same participants in the second row of Figure 3. In general, as seen in the second row of Figure 3, there are fewer associations among the variables (sparser networks), likely due to only relying only on the available data for each specific participant. This is especially true for participants who completed relatively fewer days' worth of surveys resulting in lower statistical power 2 and thus had fewer significant associations. Again, there are interesting differences across participants with certain variables being more or less central to the network or with differing numbers of associations. Note that napping is not included in Participant 65's network, which is because this participant never took a nap in the 70‐day study and thus there was no within‐person variation to model. Recognizing the likelihood of reduced power for the models of individuals who completed fewer daily surveys, we wanted to examine the robustness of such models. We did this by identifying the participants who completed 69 or 70 days' worth of surveys (n = 9) and reran the mlvar model while only including a random sample of their surveys and then correlated the resulting idiographic model with the original model that included all of their data. This represents an edge‐weight rank order similarity of the two networks. We did this for each of the nine high‐completer participants at 10, 20, 30, and 40 surveys (36 total models). One participant ended up as a relative outlier whose correlation only reached .40, even at 40 surveys. As expected, for the other participants as a group, their average correlation between their subsample idiographic model and full‐data idiographic model increased with additional days (10 days: r average  = .63, r SD  = .13; 20 days: r average  = .69, r SD  = .17; 30 days: r average  = .79, r SD  = .07; 40 days: r average  = .81, r SD  = .10).

DISCUSSION

This study assessed the multilevel clustering of lifestyle factors and aspects of emotional health in daily life in a non‐clinical sample of adults living in the United States. The models revealed evidence of clustering in the whole sample at the between‐person, within‐person contemporaneous, and within‐person temporal levels of analysis. Further, using up to 70 days' worth of data per person, a major goal of the study was to explore personalized networks of lifestyle factors and both positive and negative affect. Such models revealed important between‐person differences in how these factors covary in daily life.

Focusing first on the most novel research question of this project, it is apparent that with intensive enough data, idiographic models of lifestyle and affect clustering can be obtained. As seen in the example plots, individuals demonstrate wide variance in the strength and even direction of associations among lifestyle factors and emotional health. Such variation may be explained by a variety of factors, including differences in values, goals, self‐efficacy, attitudes, and compensatory health beliefs (Epton et al., 2017; Godin & Kok, 1996; Knäuper et al., 2004). Further, people may vary in the degree to which they have developed psychological resources and momentum, which can promote upward spirals of positive behaviors and feelings (Fredrickson, 2013; Iso‐Ahola & Dotson, 2016). From an applied perspective, the potential bidirectional associations of lifestyle and emotional health may be significantly different for each person and suggest a theoretical and empirical basis for testing individualized intervention approaches. For example, certain behaviors might result in positive affect for some participants, but not others, indicating that a broaden‐and‐build process might be triggered by different behaviors for different people. Although this study provides a proof‐of‐concept for how these processes might play out, future work must establish whether such intensive assessment is feasible and provides added value in clinical practice. Further, personalized lifestyle interventions that are informed by individualized network models may result in improved health outcomes by targeting highly central nodes and by increasing motivation and buy‐in from participants who know their treatment is tailored to their specific presentation (Li et al., 2024). Levinson et al. (2026), who are at the forefront of network‐informed personalized treatments, have found initial indications that such an approach can work for eating disorders. Translating this work into a network‐informed personalized lifestyle behavior change intervention may have high public health relevance for the treatment and prevention of physical and mental health disorders (Anderson, 2025).

Regarding the statistical approaches to modeling person‐specific associations, each may have unique applications for improving lifestyle and health. The person‐specific graphicalVAR models relied only on the participant's data, such that if an individual has consistent non‐engagement in a behavior (e.g. they never spend time in nature), then that variable will have no variance for that person and cannot be included in the model. This holds true even for lifestyle factors which the person engages in but does so at the same exact level every day (e.g. exactly two cups of coffee every day with no variation), as there would be no day‐to‐day variation to model. According to Epskamp, van Borkulo, et al.'s (2018) simulations, the graphicalVAR approach in this study would likely have benefitted from more data to achieve accurate estimates. However, the benefit of the graphicalVAR models is that they produce more conservative daily associations for a given participant. On the other hand, the individual network models estimated in the mlVAR package borrow information from the within‐person fixed effects in the whole sample. In this way, if a within‐person association is strong across the sample (nature engagement being positively associated with physical activity), then even if the participant has limited variation in one of those variables, it is possible that their individual network may still have such an association, albeit attenuated. Imagine a client is presenting with an extremely limited behavioral repertoire, with consistent non‐engagement in various healthy lifestyle behaviors. A graphicalVAR network model for that participant might only have a few relevant lifestyle factors included. However, the tentative associations borrowed from the whole sample within‐person network in an mlVAR model might provide helpful clinical hypotheses about which lifestyle behaviors might be worth targeting to improve emotional health. However, because the current sample is non‐clinical and convenience‐based, these hypotheses currently serve as a foundation for future research in clinical populations rather than immediate clinical guidelines. Thus, the graphicalVAR network represents a more conservative person‐specific approach, whereas the mlVAR network balances idiographic differences with the sample‐wide estimates. Considering the comparison of networks with differing numbers of survey data to full‐data models, although there are no recommended cutoffs for correlations which represent equivalent networks, having greater than 20 data points is probably a starting point, with more data producing more accurate networks.

In line with previous research, day‐to‐day variation of health factors were associated with other health factors, controlling for between‐person differences. Network models are by and large an exploratory analysis, and although we did see associations in the directions we expected for some variables, we did not for others. In general, this within‐person clustering of lifestyle factors was robust to the inclusion of positive and negative affect in the model and may have important implications for promoting lifestyle change. On average, a poorer than average day in one behavior may be predictive of an overall downward spiral in the overall lifestyle profile, whereas a greater than average day in one behavior may be associated with an upward spiral of other healthy behaviors, potentially due to increased resilience (Zhang et al., 2022). Take, for example, the connections among nature engagement, physical activity, social interaction, fruit and vegetable intake, and positive affect in Figure 2—these positive factors may be self‐reinforcing in a given day and help support greater daily health and well‐being. Providers attempting to support change in one behavior might need to consider whether a simultaneous emphasis on changing the other connected behaviors is warranted (Silva et al., 2024; Wilson et al., 2015). Although the novel person‐specific networks may useful for designing personalized interventions, the sample‐wide within‐person contemporaneous network models may still be a helpful tool when personalized data collection is not feasible, providing the best guess at what lifestyle factors and aspects of emotional health might covary in daily life. Further, it is interesting to examine the ways in which individuals' networks vary from the sample, as it prompts questions and hypotheses about their idiosyncratic patterns.

At the between‐person level, we see evidence that sleep factors such as sleep duration, sleep hours, next‐day napping, and next‐day caffeine use cluster together, which has been widely demonstrated by previous research (Regestein et al., 2010; Watson et al., 2016), whereas healthy lifestyle behaviors including fruit and vegetable intake, physical activity, and nature engagement also cluster together (Hobbs et al., 2015; Poortinga, 2007). These findings contribute information about patterns of associations for the aggregated lifestyle factors, which are not available using within‐person associations. For example, why do people who spend more time in nature eat more fruit and vegetables? There is no within‐person association between the two variables, but the between‐person association indicates some potential underlying stable characteristic, such as affinity toward connection to the natural world that might have been missed in within‐person analyses. One strength of the between‐person associations in this study was that they were based off averages up to 70 days' worth of data per participant, rather than retrospective aggregated assessments, which can reduce deficits and biases due to recall. One way in which such findings might have practical relevance is that the clustering of lifestyle factors (for good or ill) might be part of what contributes to health disparities for underserved and minoritized populations. Narrow behavioral repertoires, minimal resources, and a lack of psychological momentum may create downward spirals of poor lifestyles and emotional distress (Herz et al., 2020; van Cappellen et al., 2018). This process may be driven by and interact with environmental conditions. For example, previous research indicates that institutions that support poor lifestyle (fast food restaurants, liquor stores, and smoke shops) tend to cluster together geographically in low resource areas (Macdonald et al., 2018). Further, most jobs are sedentary in nature (Burnette et al., 2023) and over 100 million people nationwide lack access to a park or walking trail within a 10‐min walk of their home (Trust for Public Land, n.d.). Thus, the patterns of clustering demonstrated in this study may provide a lens through which future research could examine how social determinants of health might drive downward spirals in underserved populations.

Although recent research has shown some self‐rated improvement in various healthy lifestyle behaviors, continued improvement in lifestyle profiles is needed to improve overall health (Li et al., 2023). The group level networks can highlight general tendencies or associations within populations, which can be useful for suggesting guidelines for first‐line interventions. Yet, in light of the significant idiographic variation demonstrated by this study, the use of individualized network models could guide personalized holistic lifestyle interventions, which may be especially effective for obtaining desired health outcomes (Levinson et al., 2023).

Limitations and future directions

There are important limitations to address in this study. The sample was not representative of the wider US population and was not a clinical sample, which limits the ability to generalize these findings as lifestyle behaviors are impacted by cultural and contextual factors (Jayasinghe et al., 2025; Rodgers et al., 2025). Future studies in this area of research should examine individualized network models of lifestyle and emotional health in clinical samples, to begin to explore the feasibility of such intensive measurement within clinical care. This work could examine the lower bounds of data, which are needed to adequately model idiographic associations, the impact of missingness, and differences in temporal spacing of measurement using simulation studies (e.g. Epskamp, van Borkulo, et al., 2018). The study also relied only on mostly brief, self‐report assessments of various lifestyle factors, some of which are better measured by biometric assessments (e.g. physical activity and sleep), which should be implemented in future research. Of particular note, the low within‐person reliability of the two‐item dietary variable likely attenuated the associations between diet and other variables. Future research should explore other brief dietary intake measures for daily diary research. Further, although daily assessment of lifestyle and affect provides higher resolution data than aggregated recall from cross‐sectional studies, emotions fluctuate throughout the day and the associations among lifestyle factors and affect may be better assessed by more intensive (multiple time per day) assessment that take context into account (e.g. at home, school, and work). Such analyses might support the exploration of more dynamic causal relationships among relevant behaviors and affect states (e.g. Kekäläinen et al., 2023). Finally, this study only assessed adaptive health behaviors (i.e. eating nutritious food and exercising) and not maladaptive behaviors (e.g. eating sugary foods [Knüppel et al., 2017] and sedentary behavior [Zou et al., 2024]), which may limit our understanding of how these other factors might impact given networks of associations and the models' generalizability to unhealthy lifestyle behaviors. For example, sedentary behavior may be linked to the more problematic passive types of activities (e.g. TV viewing) but may also include more active activities such as reading, which tends to promote brain functioning (Zhang et al., 2025).

Conclusion

Lifestyle factors are increasingly being recognized as important contributors to physical and emotional health. The present study identified clustering of lifestyle behaviors at the between‐person, within‐person contemporaneous (same‐day), and within‐person temporal levels. This clustering was also modeled in person‐specific models, revealing unique patterns of clustering for each participant. Understanding how these variables interact in daily life using intensive longitudinal designs expands upon findings from cross‐sectional and large‐interval longitudinal research. Further, recognizing that these associations differ across individuals is an important preliminary step to developing and testing personalized interventions to meet clients' unique needs.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

ETHICS STATEMENT

This study was approved by the University of Southern Mississippi Institutional Review Board.

Supporting information

Data S1. Supporting Information.

APHW-18-0-s001.pdf (1.2MB, pdf)

Anderson, A. R. , Ostermiller, L. , Rice, I. , & O'Rourke, C. (2026). Examining the clustering of lifestyle factors and affect in daily life: An idiographic approach. Applied Psychology: Health and Well‐Being, 18(4), e70179. 10.1111/aphw.70179

Funding information This data collection for this study was funded by the University of Southern Mississippi, and the writing and revision of the manuscript was supported by Florida State University.

ENDNOTES

1

This does result in the temporally distant association of “last night's” sleep predicting “tomorrow's” lifestyle factors in the temporal model, but this temporal sequencing was determined to be the best option for examining these associations.

2

The average number of partial correlations >.10 was 13.4 for those with equal to or less than 25 data points (n = 11), whereas it was 22.6 for 26–54 data points (n = 24) and 21.4 for those with 55 or more data points (n = 44). Taken together, this suggests that more data per participant can better capture the relevant associations in their networks and mirrors recommendations of the mlVAR package to focus on collecting more than 20 data points per person (Epskamp et al., 2024).

DATA AVAILABILITY STATEMENT

Code and data for the analyses and supplemental materials are available in the Open Science Framework (at https://osf.io/ugtqr/?view_only=333b13cc8aee47388b1a9007b0dcdf0f).

REFERENCES

  1. Anderson, A. R. (2025). Applying the lifestyle lens to population mental health science: A commentary on Dodge et al. (2024). American Psychologist, 80(5), 835–837. 10.1037/amp0001536 [DOI] [PubMed] [Google Scholar]
  2. Anderson, A. R. , & Fowers, B. J. (2020). Lifestyle behaviors, psychological distress, and well‐being: A daily diary study. Social Science & Medicine, 263, 113263. 10.1016/j.socscimed.2020.113263 [DOI] [PubMed] [Google Scholar]
  3. Anderson, A. R. , Kurz, A. S. , Szabo, Y. Z. , McGuire, A. P. , & Frankfurt, S. B. (2022). Exploring the longitudinal clustering of lifestyle behaviors, social determinants of health, and depression. Journal of Health Psychology (Sage UK: London, England), 27, 2922–2935. 10.1177/13591053211072685 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Blanchard, L. , Conway‐Moore, K. , Aguiar, A. , Önal, F. , Rutter, H. , Helleve, A. , Nwosu, E. , Falcone, J. , Savona, N. , Boyland, E. , & Knai, C. (2023). Associations between social media, adolescent mental health, and diet: A systematic review. Obesity Reviews, 24(S2), e13631. 10.1111/obr.13631 [DOI] [PubMed] [Google Scholar]
  5. Bootsma, T. I. , Schellekens, M. P. J. , van Woezik, R. A. M. , Slatman, J. , & van der Lee, M. L. (2022). Using smartphone‐based ecological momentary assessment and personalized feedback for patients with chronic cancer‐related fatigue: A proof‐of‐concept study. Internet Interventions, 30, 100568. 10.1016/j.invent.2022.100568 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Borsboom, D. , Deserno, M. K. , Rhemtulla, M. , Epskamp, S. , Fried, E. I. , McNally, R. J. , Robinaugh, D. J. , Perugini, M. , Dalege, J. , Costantini, G. , Isvoranu, A.‐M. , Wysocki, A. C. , van Borkulo, C. D. , Bork, R. , & Waldorp, L. J. (2021). Network analysis of multivariate data in psychological science. Nature Reviews Methods Primers, 1(1), 58. 10.1038/s43586-021-00055-w [DOI] [Google Scholar]
  7. Brunsø, K. , Scholderer, J. , & Grunert, K. G. (2004). Closing the gap between values and behavior—A means–end theory of lifestyle. Journal of Business Research, Marketing Communications and Consumer Behavior, 57(6), 665–670. 10.1016/S0148-2963(02)00310-7 [DOI] [Google Scholar]
  8. Burnette, J. L. , Billingsley, J. , Banks, G. C. , Knouse, L. E. , Hoyt, C. L. , Pollack, J. M. , & Simon, S. (2023). A systematic review and meta‐analysis of growth mindset interventions: For whom, how, and why might such interventions work? Psychological Bulletin, 149(3–4), 174–205. 10.1037/bul0000368 [DOI] [PubMed] [Google Scholar]
  9. Cameron, D. S. , Bertenshaw, E. J. , & Sheeran, P. (2018). Positive affect and physical activity: Testing effects on goal setting, activation, prioritisation, and attainment. Psychology & Health, 33(2), 258–274. 10.1080/08870446.2017.1314477 [DOI] [PubMed] [Google Scholar]
  10. Chevance, G. , Golaszewski, N. M. , Baretta, D. , Hekler, E. B. , Larsen, B. A. , Patrick, K. , & Godino, J. (2020). Modelling multiple health behavior change with network analyses: Results from a one‐year study conducted among overweight and obese adults. Journal of Behavioral Medicine, 43(2), 254–261. 10.1007/s10865-020-00137-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Dennaoui, N. , Kolt, G. S. , Guagliano, J. M. , & George, E. S. (2024). Participation in physical activity and sport in adolescent girls from Middle Eastern backgrounds. Ethnicity & Health, 29(7), 756–773. 10.1080/13557858.2024.2376054 [DOI] [PubMed] [Google Scholar]
  12. Diener, E. , Wirtz, D. , Tov, W. , Kim‐Prieto, C. , Choi, D. , Oishi, S. , & Biswas‐Diener, R. (2010). New well‐being measures: Short scales to assess flourishing and positive and negative feelings. Social Indicators Research, 97(2), 143–156. 10.1007/s11205-009-9493-y [DOI] [Google Scholar]
  13. El‐Malahi, O. , Mohajeri, D. , Mincu, R. , Bäuerle, A. , Rothenaicher, K. , Knuschke, R. , Rammos, C. , Rassaf, T. , & Lortz, J. (2024). Beneficial impacts of physical activity on heart rate variability: A systematic review and meta‐analysis. PLoS ONE, 19(4), e0299793. 10.1371/journal.pone.0299793 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Epskamp, S. , Deserno, M. K. , & Bringmann, L. F. (2024). “mlVAR: Multi‐level vector autoregression (Version 0.5.3) [computer software].” https://github.com/SachaEpskamp/mlVAR
  15. Epskamp, S. , van Borkulo, C. D. , van der Veen, D. C. , Servaas, M. N. , Isvoranu, A.‐M. , Riese, H. , & Cramer, A. O. J. (2018). Personalized network modeling in psychopathology: The importance of contemporaneous and temporal connections. Clinical Psychological Science, 6(3), 416–427. 10.1177/2167702617744325 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Epskamp, S. , Waldorp, L. J. , Mõttus, R. , & Borsboom, D. (2018). The Gaussian graphical model in cross‐sectional and time‐series data. Multivariate Behavioral Research, 53(4), 453–480. 10.1080/00273171.2018.1454823 [DOI] [PubMed] [Google Scholar]
  17. Epton, T. , Currie, S. , & Armitage, C. J. (2017). Unique effects of setting goals on behavior change: Systematic review and meta‐analysis. Journal of Consulting and Clinical Psychology, 85(12), 1182–1198 (2017‐53491‐007). 10.1037/ccp0000260 [DOI] [PubMed] [Google Scholar]
  18. Ferdosi, M. , Isfahani, B. N. , & Kolahdozan, M. S. (2020). The relationship between personality factors, social support, and regulation with lifestyle. Evidence Based Health Policy, Management and Economics. 10.18502/jebhpme.v4i3.4164 [DOI] [Google Scholar]
  19. Firth, J. , Solmi, M. , Wootton, R. E. , Vancampfort, D. , Schuch, F. B. , Hoare, E. , Gilbody, S. , Torous, J. , Teasdale, S. B. , Jackson, S. E. , Smith, L. , Eaton, M. , Jacka, F. N. , Veronese, N. , Marx, W. , Ashdown‐Franks, G. , Siskind, D. , Sarris, J. , Rosenbaum, S. , … Stubbs, B. (2020). A meta‐review of “lifestyle psychiatry”: The role of exercise, smoking, diet and sleep in the prevention and treatment of mental disorders. World Psychiatry, 19(3), 360–380. 10.1002/wps.20773 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Fisher, A. J. , Bosley, H. G. , Fernandez, K. C. , Reeves, J. W. , Soyster, P. D. , Diamond, A. E. , & Barkin, J. (2019). Open trial of a personalized modular treatment for mood and anxiety. Behaviour Research and Therapy, 116, 69–79. 10.1016/j.brat.2019.01.010 [DOI] [PubMed] [Google Scholar]
  21. Flueckiger, L. , Lieb, R. , Meyer, A. H. , Witthauer, C. , & Mata, J. (2017). Day‐to‐day variations in health behaviors and daily functioning: Two intensive longitudinal studies. Journal of Behavioral Medicine, 40(2), 307–319. 10.1007/s10865-016-9787-x [DOI] [PubMed] [Google Scholar]
  22. Fredrickson, B. L. (2004). The broaden‐and‐build theory of positive emotions. Philosophical Transactions of the Royal Society B: Biological Sciences, 359(1449), 1367–1378. 10.1098/rstb.2004.1512 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Fredrickson, B. L. (2013). Positive emotions broaden and build. In Advances in experimental social psychology (Vol. 47) (pp. 1–53). Elsevier. 10.1016/B978-0-12-407236-7.00001-2 [DOI] [Google Scholar]
  24. Frumkin, M. R. , Piccirillo, M. L. , Beck, E. D. , Grossman, J. T. , & Rodebaugh, T. L. (2021). Feasibility and utility of idiographic models in the clinic: A pilot study. Psychotherapy Research: Journal of the Society for Psychotherapy Research, 31(4), 520–534. 10.1080/10503307.2020.1805133 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Giles‐Corti, B. , & Donovan, R. J. (2002). The relative influence of individual, social and physical environment determinants of physical activity. Social Science & Medicine, 54(12), 1793–1812. 10.1016/S0277-9536(01)00150-2 [DOI] [PubMed] [Google Scholar]
  26. Godin, G. , & Kok, G. (1996). The theory of planned behavior: A review of its applications to health‐related behaviors. American Journal of Health Promotion, 11(2), 87–98. 10.4278/0890-1171-11.2.87 [DOI] [PubMed] [Google Scholar]
  27. Godin, G. , & Shephard, R. J. (1985). A simple method to assess exercise behavior in the community. Canadian Journal of Applied Sport Sciences. Journal Canadien des Sciences Appliquees au Sport, 10(3), 141–146. [PubMed] [Google Scholar]
  28. Harris, L. M. , Vanzhula, I. A. , Cash, E. D. , Levinson, C. A. , & Smith, A. R. (2025). Using network analysis to personalize treatment for individuals with co‐occurring restrictive eating disorders and suicidality: A proof‐of‐concept study. Journal of Eating Disorders, 13(1), 156. 10.1186/s40337-025-01259-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Hartig, T. , Mitchell, R. , de Vries, S. , & Frumkin, H. (2014). Nature and health. Annual Review of Public Health, 35(1), 207–228. 10.1146/annurev-publhealth-032013-182443 [DOI] [PubMed] [Google Scholar]
  30. Herz, N. , Baror, S. , & Bar, M. (2020). Overarching states of mind. Trends in Cognitive Sciences, 24(3), 184–199. 10.1016/j.tics.2019.12.015 [DOI] [PubMed] [Google Scholar]
  31. Hobbs, M. , Pearson, N. , Foster, P. J. , & Biddle, S. J. H. (2015). Sedentary behaviour and diet across the lifespan: An updated systematic review. British Journal of Sports Medicine, 49(18), 1179–1188. 10.1136/bjsports-2014-093754 [DOI] [PubMed] [Google Scholar]
  32. Iso‐Ahola, S. E. , & Dotson, C. O. (2016). Psychological momentum—A key to continued success. Frontiers in Psychology, 7, 1328. 10.3389/fpsyg.2016.01328 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Jacobs, D. R. , Ainsworth, B. E. , Hartman, T. J. , & Leon, A. S. (1993). A simultaneous evaluation of 10 commonly used physical activity questionnaires. Medicine and Science in Sports and Exercise, 25(1), 81–91. 10.1249/00005768-199301000-00012 [DOI] [PubMed] [Google Scholar]
  34. James, D. (2004). Factors influencing food choices, dietary intake, and nutrition‐related attitudes among African Americans: Application of a culturally sensitive model. Ethnicity & Health, 9(4), 349–367. 10.1080/1355785042000285375 [DOI] [PubMed] [Google Scholar]
  35. Jayasinghe, S. , Byrne, N. M. , & Hills, A. P. (2025). Cultural influences on dietary choices. Progress in Cardiovascular Diseases, 90, 22–26. 10.1016/j.pcad.2025.02.003 [DOI] [PubMed] [Google Scholar]
  36. Jovanović, V. (2015). Beyond the PANAS: Incremental validity of the Scale of Positive and Negative Experience (SPANE) in relation to well‐being. Personality and Individual Differences, 86, 487–491. 10.1016/j.paid.2015.07.015 [DOI] [Google Scholar]
  37. Kekäläinen, T. , Luchetti, M. , Terracciano, A. , Gamaldo, A. A. , Mogle, J. , Lovett, H. H. , Brown, J. , Rantalainen, T. , Sliwinski, M. J. , & Sutin, A. R. (2023). Physical activity and cognitive function: Moment‐to‐moment and day‐to‐day associations. International Journal of Behavioral Nutrition and Physical Activity, 20(1), 137. 10.1186/s12966-023-01536-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Kievit, R. , Frankenhuis, W. E. , Waldorp, L. , & Borsboom, D. (2013). Simpson's paradox in psychological science: A practical guide. Frontiers in Psychology, 4, 00513. 10.3389/fpsyg.2013.00513 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Knäuper, B. , Rabiau, M. , Cohen, O. , & Patriciu, N. (2004). Compensatory health beliefs: Scale development and psychometric properties. Psychology & Health, 19(5), 607–624. 10.1080/0887044042000196737 [DOI] [Google Scholar]
  40. Knüppel, A. , Shipley, M. J. , Llewellyn, C. H. , & Brunner, E. J. (2017). Sugar intake from sweet food and beverages, common mental disorder and depression: Prospective findings from the Whitehall II study. Scientific Reports, 7(1), 6287. 10.1038/s41598-017-05649-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Lai, M. H. C. (2021). Composite reliability of multilevel data: It's about observed scores and construct meanings. Psychological Methods, 26(1), 90–102. 10.1037/met0000287 [DOI] [PubMed] [Google Scholar]
  42. Latkin, C. A. , & Knowlton, A. R. (2015). Social network assessments and interventions for health behavior change: A critical review. Behavioral Medicine, 41(3), 90–97. 10.1080/08964289.2015.1034645 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Levinson, C. A. , Slipetz, L. , Henry, T. , Crumby, E. , & Pennesi, J. L. (2026). What makes personalized treatment work? Mechanisms of change in personalized treatment for eating disorders: A proof‐of‐concept study. Behavior Therapy, 57(1), 50–62. 10.1016/j.beth.2025.04.003 [DOI] [PubMed] [Google Scholar]
  44. Levinson, C. A. , Vanzhula, I. A. , Smith, T. W. , & Stice, E. (2020). Group and longitudinal intra‐individual networks of eating disorder symptoms in adolescents and young adults at‐risk for an eating disorder. Behaviour Research and Therapy, 135, 103731. 10.1016/j.brat.2020.103731 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Levinson, C. A. , Williams, B. M. , Christian, C. , Hunt, R. A. , Keshishian, A. C. , Brosof, L. C. , Vanzhula, I. A. , Davis, G. G. , Brown, M. L. , Bridges‐Curry, Z. , Sandoval‐Araujo, L. E. , & Ralph‐Nearman, C. (2023). Personalizing eating disorder treatment using idiographic models: An open series trial. Journal of Consulting and Clinical Psychology, 91(1), 14–28. 10.1037/ccp0000785 [DOI] [PubMed] [Google Scholar]
  46. Li, E. , Kealy, D. , Aafjes‐van Doorn, K. , McCollum, J. , Curtis, J. T. , Luo, X. , & Silberschatz, G. (2024). “It felt like I was being tailored to the treatment rather than the treatment being tailored to me”: Patient experiences of helpful and unhelpful psychotherapy. Psychotherapy Research, 1–15, 695–709. 10.1080/10503307.2024.2360448 [DOI] [PubMed] [Google Scholar]
  47. Li, Y. , Schoufour, J. , Wang, D. D. , Dhana, K. , Pan, A. , Liu, X. , Song, M. , Liu, G. , Shin, H. J. , Sun, Q. , Al‐Shaar, L. , Wang, M. , Rimm, E. B. , Hertzmark, E. , Stampfer, M. J. , Willett, W. C. , Franco, O. H. , & Hu, F. B. (2020). Healthy lifestyle and life expectancy free of cancer, cardiovascular disease, and type 2 diabetes: Prospective cohort study. BMJ, 368, l6669. 10.1136/bmj.l6669 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Li, Y. , Xia, P.‐F. , Geng, T.‐T. , Tu, Z.‐Z. , Zhang, Y.‐B. , Yu, H.‐C. , Zhang, J.‐J. , Guo, K. , Yang, K. , Liu, G. , Shan, Z. , & Pan, A. (2023). Trends in self‐reported adherence to healthy lifestyle behaviors among US adults, 1999 to March 2020. JAMA Network Open, 6(7), e2323584. 10.1001/jamanetworkopen.2023.23584 [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Liu, H. , Lafferty, J. , & Wasserman, L. (2009). The nonparanormal: Semiparametric estimation of high dimensional undirected graphs. Journal of Machine Learning Research, 10, 2295–2328. [PMC free article] [PubMed] [Google Scholar]
  50. Macdonald, L. , Olsen, J. R. , Shortt, N. K. , & Ellaway, A. (2018). Do ‘environmental bads’ such as alcohol, fast food, tobacco, and gambling outlets cluster and co‐locate in more deprived areas in Glasgow City, Scotland? Health & Place, 51, 224–231. 10.1016/j.healthplace.2018.04.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Meader, N. , King, K. , Moe‐Byrne, T. , Wright, K. , Graham, H. , Petticrew, M. , Power, C. , White, M. , & Sowden, A. J. (2016). A systematic review on the clustering and co‐occurrence of multiple risk behaviours. BMC Public Health, 16, 657. 10.1186/s12889-016-3373-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Mittleman, M. A. , Maclure, M. , Tofler, G. H. , Sherwood, J. B. , Goldberg, R. J. , & Muller, J. E. (1993). Triggering of acute myocardial infarction by heavy physical exertion—Protection against triggering by regular exertion. New England Journal of Medicine, 329(23), 1677–1683. 10.1056/NEJM199312023292301 [DOI] [PubMed] [Google Scholar]
  53. Nudelman, G. , Kalish, Y. , & Shiloh, S. (2019). The centrality of health behaviours: A network analytic approach. British Journal of Health Psychology, 24(1), 215–236. 10.1111/bjhp.12350 [DOI] [PubMed] [Google Scholar]
  54. Olano, H. A. , Kachan, D. , Tannenbaum, S. L. , Mehta, A. , Annane, D. , & Lee, D. J. (2015). Engagement in mindfulness practices by U.S. adults: Sociodemographic barriers. The Journal of Alternative and Complementary Medicine, 21(2), 100–102. 10.1089/acm.2014.0269 [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Palan, S. , & Schitter, C. (2018). Prolific.ac—A subject pool for online experiments. Journal of Behavioral and Experimental Finance, 17, 22–27. 10.1016/j.jbef.2017.12.004 [DOI] [Google Scholar]
  56. Poortinga, W. (2007). The prevalence and clustering of four major lifestyle risk factors in an English adult population. Preventive Medicine, 44(2), 124–128. 10.1016/j.ypmed.2006.10.006 [DOI] [PubMed] [Google Scholar]
  57. Regestein, Q. , Natarajan, V. , Pavlova, M. , Kawasaki, S. , Gleason, R. , & Koff, E. (2010). Sleep debt and depression in female college students. Psychiatry Research, 176(1), 34–39. 10.1016/j.psychres.2008.11.006 [DOI] [PubMed] [Google Scholar]
  58. Rodgers, M. , South, E. , Harden, M. , Whitehead, M. , & Sowden, A. (2025). Contextual factors in systematic reviews: Understanding public health interventions in low socioeconomic status and disadvantaged populations. Archives of Public Health, 83(1), 153. 10.1186/s13690-025-01644-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Saka, B. , & Yildirim, E. (2024). The effect of mood on risk taking: A systematic review. Current Psychology, 43(37), 29333–29345. 10.1007/s12144-024-06585-2 [DOI] [Google Scholar]
  60. Sancho‐Domingo, C. , Carballo, J. L. , Coloma‐Carmona, A. , & Buysse, D. J. (2021). Brief version of the Pittsburgh Sleep Quality Index (B‐PSQI) and measurement invariance across gender and age in a population‐based sample. Psychological Assessment, 33(2), 111–121. 10.1037/pas0000959 [DOI] [PubMed] [Google Scholar]
  61. Scholten, S. , Lischetzke, T. , & Glombiewski, J. A. (2022). Integrating theory‐based and data‐driven methods to case conceptualization: A functional analysis approach with ecological momentary assessment. Psychotherapy Research, 32(1), 52–64. 10.1080/10503307.2021.1916639 [DOI] [PubMed] [Google Scholar]
  62. Silva, C. C. , Presseau, J. , van Allen, Z. , Schenk, P. M. , Moreto, M. , Dinsmore, J. , & Marques, M. M. (2024). Effectiveness of interventions for changing more than one behavior at a time to manage chronic conditions: A systematic review and meta‐analysis. Annals of Behavioral Medicine, 58(6), 432–444. 10.1093/abm/kaae021 [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Parks and an Equitable Recovery. (n.d.). “Trust for public land . ”. Retrieved December 9, 2024, https://www.tpl.org/parks-and-an-equitable-recovery-parkscore-report
  64. van Cappellen, P. , Rice, E. L. , Catalino, L. I. , & Fredrickson, B. L. (2018). Positive affective processes underlie positive health behaviour change. Psychology & Health, 33(1), 77–97. 10.1080/08870446.2017.1320798 [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. van Os, J. , Verhagen, S. , Marsman, A. , Peeters, F. , Bak, M. , Marcelis, M. , Drukker, M. , Reininghaus, U. , Jacobs, N. , Lataster, T. , Simons, C. , Lousberg, R. , Gülöksüz, S. , Leue, C. , Groot, P. C. , Viechtbauer, W. , & Delespaul, P. (2017). The experience sampling method as an mHealth tool to support self‐monitoring, self‐insight, and personalized health care in clinical practice. Depression and Anxiety, 34(6), 481–493. 10.1002/da.22647 [DOI] [PubMed] [Google Scholar]
  66. Walsh, R. (2011). Lifestyle and mental health. American Psychologist, 66(7), 579–592 (2011‐01021‐001). 10.1037/a0021769 [DOI] [PubMed] [Google Scholar]
  67. Watson, E. , Coates, A. , Kohler, M. , & Banks, S. (2016). Caffeine consumption and sleep quality in Australian adults. Nutrients, 8(8), 479. 10.3390/nu8080479 [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. Wilson, K. , Senay, I. , Durantini, M. , Sánchez, F. , Hennessy, M. , Spring, B. , & Albarracín, D. (2015). When it comes to lifestyle recommendations, more is sometimes less: A meta‐analysis of theoretical assumptions underlying the effectiveness of interventions promoting multiple behavior domain change. Psychological Bulletin, 141(2), 474–509 (2014‐56564‐001). 10.1037/a0038295 [DOI] [PMC free article] [PubMed] [Google Scholar]
  69. Zhang, Y. , Pan, X.‐F. , Chen, J. , Xia, L. , Cao, A. , Zhang, Y. , Wang, J. , Li, H. , Yang, K. , Guo, K. , He, M. , & Pan, A. (2020). Combined lifestyle factors and risk of incident type 2 diabetes and prognosis among individuals with type 2 diabetes: A systematic review and meta‐analysis of prospective cohort studies. Diabetologia, 63(1), 21–33. 10.1007/s00125-019-04985-9 [DOI] [PubMed] [Google Scholar]
  70. Zhang, Z. , Chen, Y. , Yu, Q. , Li, J. , Zou, L. , Mavilidi, M. F. , Green, C. S. , Owen, N. , Hallgren, M. , Raichlen, D. , Lu, S. , Alexander, G. E. , Paas, F. , & Herold, F. (2025). A neurobiological taxonomy of sedentary behavior for brain health. Trends in Neurosciences, 48(11), 853–864. 10.1016/j.tins.2025.09.002 [DOI] [PubMed] [Google Scholar]
  71. Zhang, Z. , Wang, T. , Kuang, J. , Herold, F. , Ludyga, S. , Li, J. , Hall, D. L. , Taylor, A. , Healy, S. , Yeung, A. S. , Kramer, A. F. , & Zou, L. (2022). The roles of exercise tolerance and resilience in the effect of physical activity on emotional states among college students. International Journal of Clinical and Health Psychology, 22(3), 100312. 10.1016/j.ijchp.2022.100312 [DOI] [PMC free article] [PubMed] [Google Scholar]
  72. Zou, L. , Herold, F. , Cheval, B. , Wheeler, M. J. , Pindus, D. M. , Erickson, K. I. , Raichlen, D. A. , Alexander, G. E. , Müller, N. G. , Dunstan, D. W. , Kramer, A. F. , Hillman, C. H. , Hallgren, M. , Ekelund, U. , Maltagliati, S. , & Owen, N. (2024). Sedentary behavior and lifespan brain health. Trends in Cognitive Sciences, 28(4), 369–382. 10.1016/j.tics.2024.02.003 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Data S1. Supporting Information.

APHW-18-0-s001.pdf (1.2MB, pdf)

Data Availability Statement

Code and data for the analyses and supplemental materials are available in the Open Science Framework (at https://osf.io/ugtqr/?view_only=333b13cc8aee47388b1a9007b0dcdf0f).

Articles from Applied Psychology. Health and Well-Being are provided here courtesy of Wiley

RESOURCES