my continuous dependent variable, which measures the percentage of women in ministerial cabinet positions, is bounded between 0 and 1. My initial regression analysis has highlighted that my dependent variable has bimodal residuals (near 20 and 80 percent). This is obviously an issue, but I'm not sure how to fix it. I considered turning the dependent variable into a dummy variable (existence of women in cabinet/no women); however, it just wouldn't make sense for the purposes of this project, since I need to measure different levels women in these positions and only four instances have the value "0".
I have considered transforming the variable into an ordinal measurement, which would probably between 4-5 categories; however, I'm worried that this may mess with my analysis. Is it possible to treat this variable as an ordinal measurement as it is?
I have considered transforming the variable into an ordinal measurement, which would probably between 4-5 categories; however, I'm worried that this may mess with my analysis. Is it possible to treat this variable as an ordinal measurement as it is?