Since the study’s interest was to model the predictors for parental involvement in
homework help (≥ 3 days in an average week that a child is helped with his/her
homework), a categorical dichotomous outcome, a logit analysis was used to model the
effects of household income, gender, and race on the probability of being considered
having parental involvement in homework.
Since parental involvement was not a continuous measure, it makes sense that the
study’s dependent variable (DV) be a logit (log odds). According to Warner (2008)
binary logistic regression does not require as restrictive assumptions as other analyses
(multiple linear regression and discriminant analysis). The model assumptions of binary
logistic regression are as follows: (1) A dichotomous outcome (DV) variable (usually
coded “1” and “0”), and (2) Scores on the outcome variable must be statistically
independent of each other (Warner, 2008, p. 932). These two assumptions were met in
the study.