Can I use imputed variable “INCIMP5” as an income variable to give information about total family income? I like to use this variable because of its non-interval category of income. I categorized into five categories:
$24999 and less
$25000-$49999
$50000-$74999
$75000-$99999
$100000 and over
I am wondering whether there are other standard ways to use an imputed variable in the regression analysis. I went thoroughly to the variables description, universe, weights, and so forth. Still, I would like to be clear before starting my analysis. Thank you for your help.
This is a good question. In general the use of imputed income values is a debated topic in empirical social science. On the one hand, ignoring all non-responses from your sample may limit the representative nature of a random sample. This is because it is quite likely that those who respond to income questions are different than those who do not respond. On the other hand, many researchers are skeptical about the validity of the methods used to impute values. My general suggestion in situations like these is to do both. That is, perform your analysis both with and without the imputed values and see if anything changes substantially. This will give you a good idea of how much these imputed values influence your analysis.