Table 2 Bootstrapped estimation results.

From: The impact of insufficient sleep on dietary choices and physical activity behaviors: evidence from a randomized cross-over trial

 

Food auction bids

Beverage auction bids

Coef (st error) [ηp2]

Coef (st error) [ηp2]

(1)

(2)

 

(3)

(4)

Age

0.037

0.037

Age

0.021

0.020

(0.015)

[0.010]

(0.016)

[0.002]

(0.018)

[0.000]

(0.023)

[0.001]

Female (= 1)

–0.057

–0.058

Female (= 1)

–0.043

–0.043

(0.112)

[0.000]

(0.156)

[0.001]

(0.080)

[0.007]

(0.074)

[0.002]

Minority (= 1)

–0.050

–0.050

Minority (= 1)

0.029

0.034

(0.118)

[0.000]

(0.138)

[0.001]

(0.127)

0.000]

(0.130)

[0.001]

MEQ

–0.017

–0.017

MEQ

–0.020

–0.019

(0.013)

[0.006]

(0.021)

0.001]

(0.014)

[0.000]

(0.025)

[0.005]

SR

0.048

0.049

SR

0.047

0.060

(0.053)

[0.000]

(0.079)

[0.000]

(0.062)

[0.001]

(0.086)

[0.000]

HealthDif

0.019

 

HealthDif

0.018

 

(0.017)

[0.002]

 

(0.021)

[0.003]

 

TasteDif

0.242

 

TasteDif

0.251

 

(0.016)

[0.165]

 

(0.015)

[0.199]

 

HealthDifSR

0.006

 

HealthDifSR

–0.021

 

(0.015)

[0.001]

 

(0.018)

[0.000]

 

TastedifSR

0.026

 

TastedifSR

0.002

 

(0.016)

[0.002]

 

(0.015)

[0.000]

 

CalorieDif

 

0.001

CalorieDif

 

0.003

 

(0.001)

[0.002]

 

(0.001)

[0.022]

CalDifSR

 

0.000

CalDifSR

 

0.000

 

(0.001)

[0.000]

 

(0.001)

[0.000]

  1. N = 1416 observations per random effects GLS regression (n = 118 participants, 12 observations per participant in each panel data set (6 unique item comparisons per food and per beverage auction, administered twice—once at the end of each treatment week) with robust standard errors (clustered at participant level) for each regression. Partial eta-squared effect sizes (rounded to three decimals) shown are from comparable OLS regressions that yield similar estimates on coefficient sign and significance. Coefficient estimates and standard errors above are the average estimate across 1000 bootstrapped regressions, with a randomly selected (with replacement) set of unique auction trials for each regression (i.e., either auction C→D or D→C, but not both, etc., for each pairwise comparison of snacks and beverages). Similar results are found when using alternative continuous sleep level controls in place of the binary SR indicator variable, as well as to using an IPW correction for sample selection (see Appendix Tables A5, A6, and A7).
Back to article page