"prefer not to say" survey answers

[Doordarshini Heerah]

Hi
I am coding survey responses for stata analysis. I have many answers which are "prefer not to say", e.g. for questions "what is your income range?". But I do not know how to analyse them?

[Clyde Schechter]

It depends on your research question. Sometimes those who prefer not to say should be treated as a separate category in the analysis, and sometimes it is best to treat it as a non-response (missing value). For simple descriptive statistics, it is best to treat it as a separate category and report how many people selected that response option. For associations (whether bivariate or multivariate) with other variables, you have to decide, on scientific, not statistical, grounds whether the people who don't want to disclose their income are likely to systematically differ on that other variable from those who do.

If so, it should be a separate category. Note that in that case, you cannot treat the income variable as ordinal because you cannot order "prefer not to say" with respect to the other income range categories.

If not, you can recode it to missing value for those analyses.

[Doordarshini Heerah]

thanks