Come, we included the number of adults in the household and the number of children in the household. Household income clearly is an important determinant of food sufficiency, but so also is household composition. Previous research found that adults in a food insufficient household will go without in order for their children to have meals [35]. Also included is whether or not the individual’s spouse/ partner is present in the household, with the reference group as no spouse/partner in the household. Finally, an indicator of whether or not the individual owns the home is included. The Current Population Survey (CPS) variable HUFAMINC was used in the logit estimation in order to analyze a range of income levels. (All ATUS respondents were previously in the Current Population Survey and sampling for the ATUS was done after respondents’ final outrotation from the CPS. The ATUS interview is usually 2? AZD-8055MedChemExpress AZD-8055 months after the final CPS interview.) In the descriptive statistics and simulated results sections, the ATUS Eating Health Module variable EEINCOME1 was used to identify individuals in households with incomes greater than 185 percent of the poverty threshold or less than 185 percent of the poverty threshold. The advantage of HUFAMINC is that it has 16 income categories, but it is collected in the first month of the CPS, making it more than 16 months old when the respondent is surveyed for the ATUS. The advantage of the EEINCOME1 variable is that it is current with thePLOS ONE | DOI:10.1371/journal.pone.0158422 July 13,10 /SNAP Benefit Cycletime diary, although it does not have the detail of HUFAMINC. We use both income measures in our analysis to take advantage of each measure’s strength. Individual characteristics. Included are a standard group of demographic and labor force characteristics: gender (indicator for female); age (age in years, indicator for teens age 15?9, and indicator for seniors age 65 or over); indicator for disabled; education level (high school graduate, some college, college degree or advanced degree); race/ethnicity (African American, Asian, Hispanic); and employment status (employed, retired). The reference group is then male, age 20?4 years old, not disabled, has less than a high school diploma, is white, non-Hispanic, and is not employed and not retired. Region. As there may be regional effects, we included indicators for metropolitan/nonmetropolitan residence (metro, with nonmetro as the reference group) and for Census region (West, South, and Northeast, with Midwest as the reference group). The resulting model is a logistic regression on the likelihood of no primary eating/drinking and no secondary eating, as explained by SNAP participation, days since issuance and an interaction term, along with controls for day of week, season, year, and household, personal, and geographic factors. Using the estimated model and the estimated means of the model variables for each group (see S2 Appendix), we simulated a full benefit cycle month of daily GDC-0084 biological activity probability of not reporting any eating occurrences.ResultsResults of the estimated logit model of the probability of not eating over the day are in Table 2. Regardless of the point in the benefit cycle, being a SNAP participant lowers the likelihood of no eating over the day (coefficient is -1.1027 and significant, odds ratio is 0.332), and the log of the days since benefit issuance, regardless of SNAP participation status, appears to lower the likelihood (-0.1435 coefficient, 0.86.Come, we included the number of adults in the household and the number of children in the household. Household income clearly is an important determinant of food sufficiency, but so also is household composition. Previous research found that adults in a food insufficient household will go without in order for their children to have meals [35]. Also included is whether or not the individual’s spouse/ partner is present in the household, with the reference group as no spouse/partner in the household. Finally, an indicator of whether or not the individual owns the home is included. The Current Population Survey (CPS) variable HUFAMINC was used in the logit estimation in order to analyze a range of income levels. (All ATUS respondents were previously in the Current Population Survey and sampling for the ATUS was done after respondents’ final outrotation from the CPS. The ATUS interview is usually 2? months after the final CPS interview.) In the descriptive statistics and simulated results sections, the ATUS Eating Health Module variable EEINCOME1 was used to identify individuals in households with incomes greater than 185 percent of the poverty threshold or less than 185 percent of the poverty threshold. The advantage of HUFAMINC is that it has 16 income categories, but it is collected in the first month of the CPS, making it more than 16 months old when the respondent is surveyed for the ATUS. The advantage of the EEINCOME1 variable is that it is current with thePLOS ONE | DOI:10.1371/journal.pone.0158422 July 13,10 /SNAP Benefit Cycletime diary, although it does not have the detail of HUFAMINC. We use both income measures in our analysis to take advantage of each measure’s strength. Individual characteristics. Included are a standard group of demographic and labor force characteristics: gender (indicator for female); age (age in years, indicator for teens age 15?9, and indicator for seniors age 65 or over); indicator for disabled; education level (high school graduate, some college, college degree or advanced degree); race/ethnicity (African American, Asian, Hispanic); and employment status (employed, retired). The reference group is then male, age 20?4 years old, not disabled, has less than a high school diploma, is white, non-Hispanic, and is not employed and not retired. Region. As there may be regional effects, we included indicators for metropolitan/nonmetropolitan residence (metro, with nonmetro as the reference group) and for Census region (West, South, and Northeast, with Midwest as the reference group). The resulting model is a logistic regression on the likelihood of no primary eating/drinking and no secondary eating, as explained by SNAP participation, days since issuance and an interaction term, along with controls for day of week, season, year, and household, personal, and geographic factors. Using the estimated model and the estimated means of the model variables for each group (see S2 Appendix), we simulated a full benefit cycle month of daily probability of not reporting any eating occurrences.ResultsResults of the estimated logit model of the probability of not eating over the day are in Table 2. Regardless of the point in the benefit cycle, being a SNAP participant lowers the likelihood of no eating over the day (coefficient is -1.1027 and significant, odds ratio is 0.332), and the log of the days since benefit issuance, regardless of SNAP participation status, appears to lower the likelihood (-0.1435 coefficient, 0.86.