Table 3.

Within-day level: model estimates of the multilevel models with Level-1 predictors mean SG and glucose variability (CV) and dependent variable working memory.

Effect	Estimate	SE	95% CI
			LL	UL
Model 1 − mean SG
Fixed effects
Intercept	1.76	0.12	1.53	1.98
Mean SGa	0.00	0.00	−0.00	0.00
Trial Number	0.03	0.01	0.01	0.04
Concentrationb	0.01	0.00	0.01	0.02
HOMA-IRc	−0.20	0.10	−0.40	−0.00
Age	−0.02	0.01	−0.03	−0.01
Sex	0.15	0.15	−0.14	0.45
Education	0.04	0.03	−0.01	0.10
Random effects
SD (Intercept)	0.72	0.07	0.60	0.86
SD (Mean SGa)	0.00	0.00	0.00	0.01
SD (Trial Number)	0.02	0.01	0.00	0.05
SD (Concentrationb)	0.01	0.00	0.00	0.01
Model 2 − Glucose variability (CV)
Fixed effects
Intercept	1.73	0.13	1.48	1.98
Glucose variabilityd	−0.00	0.00	−0.01	0.01
Trial Number	0.03	0.01	0.01	0.04
Concentrationb	0.01	0.00	0.01	0.02
HOMA-IRc	−0.19	0.10	−0.39	0.02
Age	−0.02	0.01	−0.03	−0.01
Sex	0.17	0.15	−0.13	0.47
Education	0.05	0.03	− 0.01	0.10
Random effects
SD (Intercept)	0.75	0.08	0.62	0.92
SD (Glucose variabilityd)	0.01	0.01	0.00	0.03
SD (Trial Number)	0.02	0.01	0.00	0.04
SD (Concentrationb)	0.01	0.00	0.00	0.01

Estimates of the two multilevel models are presented as log-odds. The inverse logit function (e.g., R function plogis) was used to convert the log-odds to proportion of correct responses (WM performance) for data interpretation.

CV coefficient of variation, CI confidence interval, HOMA-IR homeostasis model assessment of insulin resistance, LL lower limit, SE standard error, SG sensor glucose, UL upper limit. Sex (0 = male, 1 = female).

N = 103.

^aMean SG was log-transformed to correct right-skewed data.

^bSelf-reported item.

^cHOMA-IR was log transformed to correct right-skewed data.

^dGlucose variability is measured by the coefficient of variation (CV) which was log-transformed to correct right-skewed data.