I am using the well-being module from the ATUS to look at gender differences in affect during childcare activities. I am confused about the formulas presented in the codebook for this type of analysis. The handbook gives instructions for creating inverse probability weights based on the frequency that respondents engage in certain activities (in this case childcare). My problem is that the probability of engaging in childcare activites differs systematically by gender so I’m not sure that I should use them for multivariate anlyses with gender as a predictor. Is the formula in the handbook only designed for getting population level averages rather than for analysis? I.e. for determining that the average woman reports x during childcare and the average man reports y during childcare? Should I just use the person-level well-being module weights to account for the probability of being in the sample?