Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2020 Sep 1.
Published in final edited form as: Psychol Addict Behav. 2019 Jun 13;33(6):567–573. doi: 10.1037/adb0000477

Further Psychometric Analysis of the 20-item Partner Interaction Questionnaire in an Adult Sample of Smokers

Augustine Osman 1, Nancy Amodei 2, R J Lamb 3
PMCID: PMC7123497  NIHMSID: NIHMS1029876  PMID: 31192622

Abstract

The 20-item Partner Interaction Questionnaire (PIQ-20) is frequently used to assess social support for adults wishing to stop smoking. Given that social support may play a significant role in quitting success, there is a need to understand the structure and psychometric properties of assessment instruments designed to measure the construct of partner support. The current study examined the psychometric properties of the PIQ-20 when used to assess the frequency of partner behaviors. The study participants included 380 adult volunteers (M age = 41, SD = 12; 58% male). To assess internal consistency, we used both the traditional coefficient alpha and the latent variable modeling composite reliability (coefficient-ρ) procedures. We conducted independent factor analytic methods to address issues of dimensionality and scoring of responses to the PIQ-20 items. Also, we used an item response theory modeling (IRT) procedure to examine the specificity of scores on the items. Reliability estimates for the PIQ-20 subscale scores were adequate (values ≥.70). The bifactor analysis supported deriving a total score for each subscale. IRT modeling demonstrated that the discrimination (a-slope) parameter for each subscale item was significantly different from zero. The majority of items were associated strongly with their respective subscales. Twelve items were identified that could be adopted as a potential short form of the PIQ-20. The PIQ-20 or short form provides an opportunity for assessing positive and negative partner support simultaneously. There is empirical support for the dimensional structures and scoring of responses for both versions of the instrument.

Keywords: partner support, PIQ-20, short form, dimensional analysis, item response theory

In a given year, most smokers make a quit attempt, though this attempt is often unsuccessful (Hatsukami, Stead, & Gupta, 2008; Ling & Glantz, 2004). Amount and quality of social support may be a determinant of success (Cohen, 2004; May & West, 2000). For example, Lawhon and colleagues (2009) found higher levels of positive support was associated with a greater likelihood of achieving abstinence. Important for long-term success, however, was continued minimization of negative support.

The Partner Interaction Questionnaire (PIQ) is the most used instrument for measuring social support for smoking cessation. Mermelstein, Lichtenstein, and McIntyre (1983) constructed a 76-item instrument assessing how frequently a smoker’s “spouse or living partner” engaged in positive and negative support. Based on a small sample, they recommended summing ratings to obtain total frequency scores for Positive and Negative subscales of the PIQ.

In 1990, Cohen and Lichtenstein developed a 20-item version (PIQ-20) in response to the length and inadequate psychometric properties of the original. The PIQ-20 is composed of two 10-item subscales “derived from a factor analysis of the original scale data from two clinics” (p. 305); item-factor loadings and factor intercorrelations, however, were not presented. In a series of analyses, they found that the ratio of positive to negative scores was a better predictor of abstinence than positive or negative scores alone. They also reported the Positive subscale score (coefficient-alpha = .89) and the Negative subscale score (.82) attained adequate reliability.

Burns, Rothman, Fu, Lindgren, and Joseph (2014) reexamined the structure of the PIQ-20 as they had concerns about its conceptual heterogeneity. They conducted exploratory principal components analysis (PCA). Scree plots and eigenvalues ≥ 1 criteria led to retaining four factors: Emotional Support (7 items); Complaints about Smoking (6 items); Instrumental Support (4 items); Critical of Smoker (2 items).

Although Burns and colleagues’ results are promising, further examination is needed. First, PCA is a data reduction tool and not a pure common factor-analytic method for evaluating specific dimensions (Costello & Osborne, 2005; Fabrigar, Wegener, MacCallum, & Strahan, 1999). Second, given current mixed findings regarding the structure of the PIQ-20, there is no clear empirical support for computing subscale scores. It is crucial to not only uncover the dimensions of a construct, but how to score these (Reise, Moore, & Haviland, 2010). Third, contemporary psychometric methods have not been used to examine the structure and psychometric properties of the PIQ-20. For example, because the traditional coefficient- a underestimates internal consistency, latent variable modeling probably should be used to assess internal consistency (Raykov, 1997; Reise et al., 2010).

This study had four aims: 1) examine distributional properties of scores of PIQ-20 items; 2) explore internal consistency of scores; 3) explore the structure of the PIQ-20 across “best practice” factor analytic methods; and 4) examine the specificity of items. A supplemental aim was to identify potential items for an even shorter form.

Method

Participants

The data used were from Lamb, Kirby, Morral, Galbicka, & Iguchi, 2010, Romanowich & Lamb, 2014 and Romanowich & Lamb, 2015. Participants enrolled in trials of contingency management for smoking could choose between studies for those interested in quitting and those uninterested. Participants included in this analysis were enrolled in studies for those interested in quitting. They met the criterion of smoking ≥15 cigarettes per day in order to reduce the number able to quit unaided. The sample included 221 males and 159 females.

Participants’ mean age was 41 years (SD =12) and 58% were male. They were mostly Caucasian (67%), non-white Hispanic (15%) or African American (10%). Most were single (37%) or married (30%). The mean breath CO level was 24 ppm. Participants smoked an average of 24 cigarettes daily. Each was asked to identify a primary source of support: 31% identified a spouse, 20% a close or best friend, 18% a family member, 9% a girlfriend, 5% a romantic partner, 5% a boyfriend, and 12% other. Lastly, 46% of married participants identified their spouses as current smokers, and 54% identified their spouses as non-smokers.

Measures and Procedure

Study protocols were approved by the University of Texas Health Science Center at San Antonio institutional review board. Following informed consent, participants completed a 30-minute questionnaire packet that included the Partner Interaction Questionnaire-20 (PIQ-20; Cohen & Lichtenstein, 1990). The PIQ-20 is a self-report measure assessing the frequency of support behaviors. It consists of two proposed 10-item clusters (Positive and Negative subscales). Although the developers asked participants to answer the questions based on the frequency of expected behavior, our participants were instructed “Please answer the following questions about the behaviors of your spouse or romantic partner based on the last month. If you do not have a spouse or romantic partner, pick the person, friend or relative who follows the progress of your quit attempt most closely.” Each behavior was rated on a 5-point Likert scale ranging from 0 (never) to 4 (very often). Cohen and Lichtenstein (1990) suggest positive and negative subscale scores can be generated by summing responses to the items in each subscale. Thus, subscale scores can range from 0 to 40. Positive and negative statements were presented in a mixed order.

Results

Distributions of Individual Item Scores

Researchers have noted that absolute values of skewness (+/−3) and kurtosis (+/−8) are preferred for judging non-normality when the sample size is ≥ 300 (Dimitrov, 2012; Kim, 2013; Kline, 2005). Means, standard deviations, absolute skewness, and kurtosis values for individual items are presented in Table 1. All absolute skewness values were < 3, and all kurtosis values < 8.

Table 1.

Descriptive Statistics for the PIQ-20 Items and Reliability Analyses (N = 380)

Partner Interaction Questionnaire-20 Item-level Descriptive Statistics
Positive Partner Items M SD Skewness Kurtosis Range
1. Help you think of substitutes for smoking. 1.92 1.24 −0.03 −0.83 0.0 −.04
2. Congratulate you for your decision to quit. 2.71 1.23 −0.76 −0.28 0.0 −.04
6. Help calm you down when feeling stressed. 2.36 1.21 −0.39 −.055 0.0 −.04
8. Express pleasure at your efforts to quit. 2.68 1.16 −0.61 −0.29 0.0 −.04
9. Celebrate your quitting with you. 2.19 1.36 −0.27 −1.07 0.0 −.04
10. Participate in an activity with you that keeps you from smoking (e.g., going for a walk instead of smoking). 1.78 1.30 0.08 −1.06 0.0 −.04
11. Help you use substitutes for cigarettes. 1.56 1.27 0.26 −1.02 0.0 −.04
15. Express confidence in your ability to quit/remain quit. 2.27 1.19 −0.31 −0.62 0.0 −.04
17. Tell you to stick with it. 2.51 1.32 −0.56 −0.73 0.0 −.04
20. Compliment you on not smoking. 2.33 1.29 −0.37 −0.77 0.0 −.04
Mean (SD) = 22.3 (9.8); range = .00 −40; Coefficient-alpha = .928 a; Coefficient-ρ = .929
Negative Partner Items
3. Talk you out of smoking a cigarette. 1.98 1.37 0.02 −1.17 0.0 −.04
4. Comment that smoking is a dirty habit. 2.18 1.52 −0.20 −1.40 0.0 −.04
5. Comment that the house smells like smoke. 1.43 1.54 0.56 −1.20 0.0 −.04
7. Comment on your lack of willpower. 1.33 1.28 0.58 −0.73 0.0 −.04
12. Express doubt about your ability to quit/stay quit. 1.25 1.23 0.61 −0.69 0.0 −.04
13. Mentioned being bothered by smoke. 1.83 1.49 0.12 −1.36 0.0 −.04
14. Refuse to let you smoke in the house. 2.09 1.84 −0.09 −1.84 0.0 −.04
16. Refuse to clean up your cigarette butts. 1.92 1.73 0.08 −1.73 0.0 −.04
18. Criticize your smoking. 1.84 1.49 0.09 −1.38 0.0 −.04
19. Asked you to quit smoking. 2.00 1.59 −0.05 −1.54 0.0 −.04
Mean (SD) = 17.8 (9.9); range = .00–40; Coefficient-alpha = .851 b; Coefficient-ρ = .862
Open in a new tab

Note. PIQ-20 = Partner Interaction Questionnaire-20, M = mean, SD = standard deviation.

a

Corrected item-total subscale correlations: range = .55−.81

b

Corrected item-total subscale correlations: range = .28−.80.

Internal Consistency Analysis and Subgroup Analysis

Cronbach coefficient-α and the latent variable coefficient-ρ were computed (Raykov, 2009). Values ≥ .70 indicate adequate internal consistency. Coefficient-ρ is typically preferred because it takes into consideration factor loadings and related measurement errors. The analysis was conducted with Mplus 8 (Muthén & Muthén, 1998–2017).

Table 1 shows the Cronbach-α and coefficient-ρ estimates. For the Positive subscale, the corrected item-total correlations ranged from .55 to .81. For the Negative subscale, these ranged from .28 to .80. Only Item 12 (“Express doubt about your ability to quit/stay quit”) from the Negative subscale was < .30. Removing it did not substantially increase the internal consistency of the remaining nine items of the Negative subscale (.85 vs. .86); thus, item 12 was retained. Coefficient-ρ also suggested subscales were reliable. Composite reliability for the Positive subscale was high, .93 (95% CI = .91; .94). The coefficient-ρ for the Negative subscale was adequate, .86 (95% CI = .84; .88).

Conventional Exploratory Factor Analysis

Because of the limited PIQ-20 factor analytic studies, we conducted a conventional common exploratory factor analysis. We used the minimum rank factor analysis with Promax rotation. Unlike maximum likelihood, this estimator provides meaningful solutions even when the matrix is positive semi-definite (ten Berge & Kiers, 1991). Both Horn’s parallel analysis and Velicer’s minimum average partial tests recommended extracting 2-factors (Hayton, Allen, & Scarpello, 2004; Horn, 1965; Velicer, 1976). For all items, only loadings ≥ .30 on one factor and ≤.29 on the other factor were considered specific. Moreover, only factors composed of ≥ four items were used in further analyses (see Fabrigar et al., 1999; Floyd & Widaman, 1995).

Table 2 shows the rotated standardized factor loadings and the discrimination parameters (see factor loading). The correlation between the Positive and Negative factor was .46, indicating moderate association. The Positive factor (43.2% variance) had 10 items, with items loading positively at ≥.30. Two items with cross-loadings were identified for this factor. Item 12 (Negative items) had a negative cross-loading (λ = −.43); Item 3 (Negative items) had a positive cross-loading (λ = .57). The Negative factor (28.9% variance) was composed of nine items with loadings of ≥.30. Item 3 had a loading of .29 (Comrey & Lee, 1992).

Table 2.

Standardized Factor Loadings and Parameter Estimates for the PIQ-20 Items

Conventional Exploratory Factor Analysis Bifactor ESEM R-programming
Factor loading Slope
Item # Question PP NP PP NP Gen PP NP Gen PP NP Slope
Positive Support Subscale
8 Express pleasure at your efforts to quit. 0.87 0.03 2.66 0.09 0.69 0.59 0.60 0.62 2.81
20 Compliment you on not smoking. 0.89 −0.03 2.43 −0.07 0.64 0.63 0.57 0.63 2.63
2 Congratulate you for your decision to quit. 0.87 −0.05 3.44 −0.19 0.66 0.55 0.55 0.61 2.37
17 Tell you to stick with it. 0.87 −0.07 2.53 −0.20 0.57 0.64 0.53 0.61 2.29
9 Celebrate your quitting with you. 0.86 −0.06 1.91 −0.13 0.64 0.53 0.53 0.60 2.25
15 Express confidence in your ability to quit/remain quit. 0.89 −0.17 1.95 −0.38 0.49 0.70 0.49 0.62 1.95
11 Help you use substitutes for cigarettes. 0.71 0.05 1.46 0.10 0.70 0.31 0.47 0.48 1.83
10 Participate in an activity with you that keeps you from smoking (e.g., going for a walk instead of smoking). 0.71 0.07 1.47 0.13 0.65 0.36 0.48 0.48 1.78
6 Help calm you down when you are feeling stressed or irritable. 0.68 −0.16 1.11 −0.25 0.46 0.35 0.33 0.45 1.22
1 Help you think of substitutes for smoking. 0.62 0.25 1.44 0.57 0.77 0.21 0.53 0.41 1.98
Negative Support Subscale
19 Asked you to quit smoking. 0.25 0.73 0.88 2.56 0.76 0.48 0.60 0.56 1.60
4 Comment that smoking is a dirty habit. 0.18 0.75 0.75 3.18 0.75 0.33 0.54 0.52 1.36
18 Criticize your smoking. 0.02 0.90 0.12 6.22 0.69 0.62 0.57 0.66 1.19
13 Mentioned being bothered by smoke. 0.06 0.82 0.20 2.56 0.63 0.57 0.54 0.59 1.18
14 Refuse to let you smoke in the house. 0.21 0.53 0.48 1.22 0.50 0.48 0.43 0.37 1.13
16 Refuse to clean up your cigarette butts. −0.07 0.62 −0.14 1.13 0.37 0.47 0.30 0.40 0.65
7 Comment on your lack of willpower. -0.19 0.74 −0.46 1.84 0.60 0.08 0.30 0.42 0.69
5 Comment that the house smells like smoke. −0.14 0.63 −0.33 1.46 0.44 0.20 0.26 0.36 0.58
12 Express doubt about your ability to quit/stay quit. −0.43 0.64 −0.81 1.21 0.34 0.01 0.00 −0.23 0.33 0.26
3 Talk you out of smoking a cigarette. 0.57 0.29 1.07 0.55 0.74 0.03 0.55 0.37 0.24 1.84
Open in a new tab

Note. PP = Positive Partner; NP = Negative Partner; Gen = General Distress; ESEM = exploratory structural equation modeling.

An advantage of using FACTOR 9.3 (Lorenzo-Seva & Ferrando, 2006–2014) in the present study is that it provides an opportunity to examine the discrimination parameters, which evaluate the strength of the relationship between each item and the target dimension. For the Positive subscale, all the items had discrimination parameters of 1.0 or higher, indicating moderate associations with this factor (Slope PP). Also, for the Negative subscale (Slope NP), all the items except one (item 3) were moderately linked with this factor.

Bifactor Exploratory Structural Equation Modeling.

First, we examined fit estimates of the oblique 4-factor solution proposed by Burns et al. (2014), using all 20 items. This model could not be fitted and was not further considered. We conducted exploratory bifactor ESEM using Mplus version 8 (Muthén & Muthén, 1998–2017) to examine alternative structures of the PIQ-20, and address concerns about scoring and interpretation. Bifactor modeling permits analyses of unique contributions of general and group-specific dimensions by constraining each item to load on a general factor (g) and a group-specific factor (s). The group-specific factors are specified to be orthogonal. The specificity of each item is judged in terms of higher loadings on s than on g (Asparouhov & Muthén, 2009; Reise et al., 2010; Reise, Scheines, Widaman, & Haviland, 2013). We specified a one-factor, a two-factor, and the bifactor model.

We considered 3 evaluative fit indexes (Browne & Cudeck, 1993; Hu & Bentler, 1999; Marsh, Hau, & Wen, 2004; Tucker & Lewis, 1973): comparative fit index (CFI ≥ .95); Tucker-Lewis Index (TLI ≥ .95); and root-mean-square error of approximation (RMSEA ≥ .05 as strong; ≥ .08 as reasonable fit). We used the robust estimator, weighted least squares estimation with mean- and variance-adjusted standard errors method (WLSMV) with orthogonal rotation.

We used the R psych package (R Development Core Team, 2011) to assess bifactor statistics: explained common variance (ECV), omega hierarchical (ωh) and omega subscale (ωs) (McDonald, 1999). The omega estimates represent measures of the reliability of latent constructs in a bifactor model when the effects of other constructs are removed (Reise et al., 2010). ECV serves as an index of unidimensionality for potentially multidimensional data. Specifically, unidimensionality is indicated if ECV ≥ .60 (Reise et al., 2013).

The one-factor model fit poorly: Χ2 (170) = 2,207.75, p < .001; CFI = 0.83, TLI = 0.81, RMSEA = .18 (90% CI = .17; .18). The 2-factor correlated model fit satisfactorily, but the large RMSEA suggested lack of generalizability: Χ2 (151) =666.63, p < .001; CFI = 0.96, TLI = 0.95, RMSEA = .10 (90% CI = .09; .10). The bifactor solution, however, provided improved fit: Χ2 (133) = 414.94, p < .001; CFI = 0.98; TLI = 0.97; RMSEA = .08 (90% CI = .07; .08).

Table 2 shows the results of the bifactor ESEM. The Positive factor is composed of 6 items (2, 8, 9, 15, 17, 20) with loadings of ≥.50. The Negative factor is composed of 2 items (13 & 18) with loadings of ≥.50. Fifteen of 20 items had loadings of ≥.50 on the general factor. Table 2 shows results from the bifactor miners procedure of R. Six of 10 Positive factor items had loadings ≥.50. The Negative factor had 4 items with loadings ≥.50. The observed ECV of the general factor was low (0.46%), suggesting PIQ-20 subscale item variances could be considered beyond the general factor by computing total frequency scores for each. The omega hierarchical index for the PIQ-20 general (ωh) factor was moderate, 0.57. After controlling for the variance due to the general factor, ωs-p was 0.49, and ωs-n was 0.48. Using the original scoring, the estimate of internal consistency (coefficient-ω with bootstrapping) was high: PIQ-20 total = 0.941 (95% CI =.918; .963); PIQ-20 Positive subscale = 0.940 (95% CI =.915; .965), and PIQ-20 Negative subscale = 0.815 (95% CI =.777; .844).

Item Response Theory (IRT) Modeling

Item Response Theory (IRT) Modeling Xcalibre 4.2 (Guyer & Thompson, 2012) was used to examine the strength (a-parameter) of associations between items within a subscale and the related subscale. We used the Samejima’s graded response model (GRM) for polytomous data to estimate item parameters (Embretson & Reise, 2013; Reise, Widaman, & Pugh, 1993). For each item within a factor, an item-slope (a) and a location or difficulty (b) parameter was estimated. Discrimination (slope) values of 0.01– 0.34 are interpreted as very low (Baker & Kim, 2004), values of 0.35 – 0.64 as low, values of 0.65 – 1.34 as moderate, values of 1.35 – 1.69 as high, and estimates of ≥1.70 as very high. We focused on the slope parameter.

We ran separate EFA analyses with items within each subscale to address unidimensionality (Reckase, 1979). Within the Positive subscale (57.6% of common variance), all item-factor loadings were significant, ranging from .59 to .91: CFI = .98, TLI = .97. Similarly, all item-factor loadings for the Negative subscale (40.7%) were significant, ranging from .40 to .91: CFI = .93, TLI = .91. Table 2 presents one a-parameter (Slope) for each item. On the Positive subscale, items with the strongest associations were Item 8 (a = 2.81) and Item 20 (a = 2.63), followed by Item 2. On the Negative subscale, items with the strongest associations were Item 3 (a = 1.84) and Item 19 (a = 1.60), followed by Item 4. The lowest fitting items were Item 12 (a = 0.26) and Item 5 (a = 0.58).

Short form PIQ-12

While the PIQ-20 2-factor solution attained good fit across multiple methods, not all the items met the expected item-factor loading and item-discrimination parameter criteria. Further, examination of item contents indicated overlap for some items (e.g., Items 1 vs. 11; Items 7 vs. 12). Thus, we used bifactor and IRT results to identify items for each factor. Items loading higher (>.50) on a specific factor, using the bifactor analysis result were identified as potential short-form items. Also, items with discrimination parameters > 0.60 (i.e., use of less stringent criteria) allowed us to consider items with a range of moderate to high discrimination levels. Next, we conducted internal consistency and the related item-total subscale correlation and confirmatory factor analyses to ensure that we were able to recover the 2-factor solution.

Items 2, 8, 9, 15, 17, & 20 were included in the PIQ-12 short-form Positive subscale. Items 4, 13, 14, 16, 18, & 19 were included in the PIQ-12 Negative subscale. These items are bolded in Table 2. Table 3 lists the items renumbered for use as a stand-alone instrument. CFA (WLSMV with Geomin rotation) of the 12 items indicated excellent fit for the 2-factor solution: chi-square = 175.97, df = 53; CFI = .99, TLI = .98, RMSEA = 0.07 (0.06, 0.09). The PIQ-12 Positive subscale standardized loadings ranged from .81 to .92. The Negative subscale standardized loadings ranged from .53 to .91. The correlation between the factor solutions was moderate at .48. Internal consistency was excellent for subscale scores of the PIQ-12. The Positive subscale score (M = 14.70, SD = 6.47; N = 380) showed evidence of good internal consistency: coefficient omega = .92 (95% CI =.91 – .94; corrected item-total subscale corrrelations, .73 – .83). Likewise, the Negative subscale score (M = 11.85, SD = 7.42; N = 380) demonstrated evidence of adequate internal consistency: coefficient omega = .85 (95% CI =.82 −.87; corrected item-total subscale correlations .47 – .80). The correlation between the PIQ-20 and the PIQ-12 Positive subscale scores was high, .96 (95% CI = .96, .97). Similarly, the correlation between the PIQ-20 and the PIQ-12 Negative subscale scores was high, .96 (95% CI = .95, .96).

Table 3.

Confirmatory Factor Analysis of the PIQ-12 Items

Standardized
Loadings
Factor 1: Positive Partner Factor 1 Factor 2 S.E.
1. Compliment you on not smoking (20) .90 .01
2. Express pleasure at your efforts to quit (8) .92 .01
5. Congratulate you for your decision to quit (2) .86 .02
7. Celebrate your quitting with you (9) .83 .02
9. Tell you to stick with it (17) .86 .02
10. Express confidence in your ability to quit/remain quit (15) .81 .02
Mean = 14.7 (SD = 6.5; range = .00–24); Coefficient-omega = .92
Factor 2: Negative Partner
3. Criticize your smoking (18) .89 .02
4. Mentioned being bothered by smoke (13) .84 .02
6. Refuse to let you smoke in the house (14) .72 .04
8. Asked you to quit (19) .91 .02
11. Comment that smoking is a dirty habit (4) .80 .03
12 Refuse to clean up your cigarette butts (16) .53 .05
Mean = 11.9 (SD = 7.4; range = .00–24); Coefficient-omega = .85
Open in a new tab

Note. S.E. = standard error, SD = standard deviation. Items from the original PIQ-20 are set in parentheses.

Discussion

To our knowledge, the current study is the first to use newer statistical strategies to address the psychometric properties of the PIQ-20. Bifactor modeling suggests that the PIQ-20 is not unidimensional and that scores on both the total (general support) and the subscales can be scored and interpreted meaningfully. IRT analysis showed that most items had moderate to very high discrimination parameters. However, unlike the Positive subscale, the Negative subscale had three extreme location items: Item 5, 7 and 12. These items had the lowest discrimination parameters, indicating that they are not strongly associated with the Negative dimension. Burns et al. (2014) included Items 7 and 12 in a domain-specific factor for their PIQ-19.

Unlike previous investigations with the PIQ-20, we were also interested in identifying individual items strongly linked to specific subscales. In both research and clinical settings, these items are the most helpful when differentiating the responses of subgroups. The CFA procedure indicated good support for the structure of this short-form, confirming that the original structure of PIQ-20 was retained. Both subscales had strong internal consistency.

The current study has limitations. Although this study focused on examining the structure of the PIQ-20, participants received instructions differing from those provided by the original developers. However, it is possible to modify instructions, but not the item contents, of an instrument without changing its psychometric properties. Second, as stated we were interested in examining the factor structure of the PIQ-20, not in how scores on scales perform over time. Given the use of a cross-sectional design, we cannot speculate on the potential performance of the PIQ-20 or PIQ-12 longitudinally. Third, we used bifactor ESEM and conventional common exploratory factor analytic procedures to examine the structure of the PIQ-20. Because previous studies reported specific exploratory solutions (Burns et al., 2014; Roski et al., 1996), it might have been useful to use CFA to evaluate the fit of alternative solutions. However, because of the limitations of principal components analysis in the validation of scores, the use of bifactor ESEM was considered appropriate (Reise, Waller, & Comrey, 2000; Widaman, 1993). Fourth, the generalizability of the study is limited in that we used smokers volunteering to participate in a contingency management study. Fifth, because we aimed to examine the structure of the PIQ-20, we did not assess the extent to which subgroups differ in their interpretation of the items. Sixth, unlike Burns et al. (2014), we did not examine other psychometric properties of the PIQ-20 or PIQ-12, including convergent and discriminant validity.

Despite these limitations, the present study supports the dimensional structure of the PIQ-20 and PIQ-12 when assessing the frequency of partner behavior. Also, this study is the first to use contemporary statistical methods to examine the structure and internal consistency of the PIQ-20 items and subscale scores.

Acknowledgments

This research was supported by NIH grant DA013304. The authors alone are responsible for the content and writing of the paper. The authors wish to thank Dr. Thomas Hoffman posthumously for his statistical contributions to earlier drafts of this manuscript.

Footnotes

The authors report no conflicts of interest.

Contributor Information

Augustine Osman, The University of Texas at San Antonio.

Nancy Amodei, University Health System, San Antonio, Texas.

R. J. Lamb, University of Texas Health Science Center at San Antonio

References

  1. Asparouhov T, & Muthén B (2009). Exploratory structural equation modeling. Structural Equation Modeling, 16, 397–438. doi: 10.1080/10705510903008204 [DOI] [Google Scholar]
  2. Baker FB, & Kim SH (Eds.). (2004). Item response theory: Parameter estimation techniques (2nd ed.). New York; Marcel Dekker. [Google Scholar]
  3. Browne MW, & Cudeck R (1993). Alternative ways of assessing model fit In Bollen KA & Long JS (Eds.), Testing structural equation models (pp. 136–162). Newbury Park, CA: Sage. [Google Scholar]
  4. Burns RJ, Rothman AJ, Fu SS, Lindgren B, & Joseph AM (2014). The relation between social support and smoking cessation: Revisiting an established measure to improve prediction. Annals of Behavioral Medicine, 47, 369–375. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Cohen S (2004). Social relationships and health. American Psychologist, 59, 676–684. [DOI] [PubMed] [Google Scholar]
  6. Cohen S, & Lichtenstein E (1990). Partner behaviors that support quitting smoking. Journal of Consulting and Clinical Psychology, 58, 304–309. [DOI] [PubMed] [Google Scholar]
  7. Comrey AL, & Lee HB (1992). A first course in factor analysis (2nd ed.), Hillsdale, New Jersey: Erlbaum. doi: 10.4324/9781315827506 [DOI] [Google Scholar]
  8. Costello AB, & Osborne JW (2005). Best practices in exploratory factor analysis: Four recommendations for getting the most from your analysis. Practical Assessment, Research & Evaluation, 10(7), 1–9. [Google Scholar]
  9. Dimitrov MD (2012). Statistical methods for validation of assessment scale data in counseling and related fields. Alexandria, VA: American Counseling Association. [Google Scholar]
  10. Embretson SE, & Reise SP (2013). Item response theory for psychologists. Psychology Press. [Google Scholar]
  11. Fabrigar LR, Wegener DT, MacCallum RC, & Strahan EJ (1999). Evaluating the use of exploratory factor analysis in psychological research. Psychological Methods, 4, 272–299. doi: 10.1037/1082-989X.4.3.272. [DOI] [Google Scholar]
  12. Floyd FJ, & Widaman KF (1995). Factor analysis in the development and refinement of clinical assessment instruments. Psychological Assessment, 7, 286–299. [Google Scholar]
  13. Ginsberg D, Hall SM, Rosinski M (1992). Partner support, psychological treatment, and nicotine gum in smoking treatment: An incremental study. International Journal of the Addictions, 27(5), 503–514. doi: 10.3109/10826089209063465 [DOI] [PubMed] [Google Scholar]
  14. Gruder CL, Mermelstein RJ, Kirkendol S, Hedeker D, Wong SC, & Schreckengost J (1993). Effects of social support and relapse prevention training as adjuncts to a televised smoking-cessation intervention. Journal of Consulting and Clinical Psychology, 61 (1), 113–120. doi: 10.1037/0022-006X.61.1.113 [DOI] [PubMed] [Google Scholar]
  15. Guyer R, & Thompson NA (2012). User’s manual for Xcalibre item response theory calibration software, Version 4.2.0. St Paul MN: Assessment Systems Corporation. [Google Scholar]
  16. Hatsukami DK, Stead LF, & Gupta PC (2008). Tobacco addiction. Lancet, 371, 2027–2038. doi: 10.1016/S0140-6736(08)60871-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Hayton JC, Allen DG, & Scarpello V (2004). Factor retention decisions in exploratory factor analysis: A tutorial on parallel analysis. Organizational Research Methods, 7, 191–205. doi: 10.1177/1094428104263675 [DOI] [Google Scholar]
  18. Horn JL (1965). A rationale and test for the number of factors in factor analysis. Psychometrika, 30, 179–185. doi: 10.1007/BF02289447 [DOI] [PubMed] [Google Scholar]
  19. Hu LT, & Bentler PM (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1–55. doi: 10.1080/10705519909540118 [DOI] [Google Scholar]
  20. Kim HY (2013). Statistical notes for clinical researchers: Assessing normal distribution (2) Using skewness and kurtosis. Restorative Dentistry & Endodontics, 38(1), 52–54. 10.5395/rde.2013.38.1.52. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Kline RB (2005). Principles and practice of structural equation modeling (2nd ed.). New York, NY: Guilford Press. [Google Scholar]
  22. Lamb RJ, Kirby KC, Morral AR, Galbicka G, & Iguchi MY (2010). Shaping smoking cessation in hard-to-treat smokers. Journal of Consulting and Clinical Psychology, 78(1), 62–71. doi: 10.1037/a0018323 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Lawhon D, Humfleet GL, Hall SM, Reus VI, & Muñoz RF (2009). Longitudinal analysis of abstinence-specific social support and smoking cessation. Health Psychology, 28, 465–472. doi: 10.1037/a0015206 [DOI] [PubMed] [Google Scholar]
  24. Ling PM, & Glantz SA (2004). Tobacco industry research in smoking cessation—Recapturing young adults and other recent quitters. Journal of General Internal Medicine, 19, 419–426. doi: 10.1111/j.1525-1497.2004.30358.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Lorenzo-Seva U, & Ferrando PJ (2006–2014). FACTOR (Version 9.3): Unrestricted factor analysis. [Computer software]. Tarragona, Spain: Author. [Google Scholar]
  26. Marsh HW, Hau KT, & Wen Z (2004). In search of golden rules: Comment on hypothesis testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing Hu and Bentler’s (1999) findings. Structural Equation Modeling, 11, 320–341. [Google Scholar]
  27. May S, & West R (2000). Do social support interventions (“buddy systems”) aid smoking cessation? A review. Tobacco Control, 9, 415–422. doi: 10.1136/tc.9.4.415 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. McDonald RP (1999). Test theory: A unified treatment. Mahwah, NJ: Erlbaum. [Google Scholar]
  29. Meredith W (1993). Measurement invariance, factor analysis and factorial invariance. Psychometrika, 58(4), 525–543. [Google Scholar]
  30. Mermelstein R, Lichtenstein E, & McIntyre K (1983). Partner support and relapse in smoking-cessation programs. Journal of Counseling and Clinical Psychology, 51(3), 465–466. doi: 10.1037/0022-006x.51.3.465 [DOI] [PubMed] [Google Scholar]
  31. Muthén LK, & Muthén B (1998–2017). Mplus user’s guide. Los Angeles, CA: Muthén & Muthén. [Google Scholar]
  32. Raykov T (1997). Estimation of composite reliability for congeneric measures. Applied Psychological Measurement, 21, 173–184. doi: 10.1177/01466216970212006 [DOI] [Google Scholar]
  33. Raykov T (2009). Evaluation of scale reliability for unidimensional measures using latent variable modeling. Measurement and Evaluation in Counseling and Development, 42(3), 223–232. doi: 10.1177/0748175609344096 [DOI] [Google Scholar]
  34. R Development Core Team. (2011). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. [Google Scholar]
  35. Reckase MD (1979). Unifactor latent trait models applied to multifactor tests: Results and implications. Journal of Educational Statistics, 4, 207–230. [Google Scholar]
  36. Reise SP, Ainsworth AT, & Haviland MG (2005). Item response theory: Fundamentals, applications, and promise in psychological research. Current directions in psychological science, 14(2), 95–101. [Google Scholar]
  37. Reise SP, Moore TM, & Haviland MG (2010). Bifactor models and rotations: Exploring the extent to which multidimensional data yield univocal scale scores. Journal of Personality Assessment, 92, 544–559. doi: 10.1080/00223891.2010.496477 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Reise SP, Scheines R, Widaman KF, & Haviland MG (2013). Multidimensionality and structural coefficient bias in structural equation modeling: A bifactor perspective. Educational and Psychological Measurement, 73, 5–26. doi: 10.1177/0013164412449831 [DOI] [Google Scholar]
  39. Reise SP, Waller NG, & Comrey AL (2000). Factor analysis and scale revision. Psychological Assessment, 12, 287–297.doi: 10.1037//1040-3590.12.3.287 [DOI] [PubMed] [Google Scholar]
  40. Reise SP, Widaman KF, & Pugh RH (1993). Confirmatory factor analysis and item response theory: Two approaches for exploring measurement invariance. Psychological Bulletin, 114, 552–566. doi: 10.1037/0033-2909.114.3.552 [DOI] [PubMed] [Google Scholar]
  41. Romanowich P, & Lamb RJ (2014). The effects of percentile versus fixed criterion schedules on smoking with equal incentive magnitude for initial abstinence. Experimental and Clinical Psychopharmacology, 22(4), 349–355. doi: 10.1037/a0036935 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Romanowich P, & Lamb RJ (2015). The effects of fixed versus escalating reinforcement schedules on smoking abstinence. Journal of Applied Behavioral Analysis, 48, 25–37. Doi: 10.1002/jaba.185 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Roski J, Schmid LA, & Lando HA (1996). Long-term associations of helpful and harmful spousal behaviors with smoking cessation. Addictive Behaviors, 21, 173–185. [DOI] [PubMed] [Google Scholar]
  44. ten Berge JMF, & Kiers HAL (1991). A numerical approach to the exact and the approximate minimum rank of a covariance matrix. Psychometrika, 56, 309–315. [Google Scholar]
  45. Tucker LR, & Lewis C (1973). A reliability coefficient for maximum likelihood factor analysis. Psychometrika, 38, 1–10. doi: 10.1007/bf02291170 [DOI] [Google Scholar]
  46. Vandenberg RJ, & Lance CE (2000). A review and synthesis of the measurement invariance literature: Suggestions, practices, and recommendations for organizational research. Organizational Research Methods, 3(1), 4–70. 10.1177/109442810031002 [DOI] [Google Scholar]
  47. Velicer WF (1976). Determining the number of components from the matrix of partial correlations. Psychometrika, 41, 321–327. doi: 10.1007/BF02293557 [DOI] [Google Scholar]
  48. Watson D, Clark LA, & Tellegen A (1988). Development and validation of brief measures of positive and negative affect: the PANAS scales. Journal of Personality & Social Psychology, 54(6),1063–1070. [DOI] [PubMed] [Google Scholar]
  49. Widaman KF (1993). Common factor analysis versus principal component analysis: Differential bias in representing model parameters? Multivariate Behavioral Research, 28, 263–311. doi: 10.1207/s15327906mbr2803_1. [DOI] [PubMed] [Google Scholar]

RESOURCES