I have a 16,000 data set with continuous dependant variables and an independent variable. I also have assigned for each observation one of about 100 distinct categories. I want to assign each category to zero or to a single non-zero value, identical across all non-zero assigned categories.

I could treat this as a non-linear optimisation problem with SSD as the objective function. But I feel sure there is a standard approach to regression/factor analysis that would address this problem.

This differs from standard factor analysis, as ex ante I don't know whether a given category will be zero or this (unknown) constant value. And if I set up the 100 categories as indicator variables and regress, they will all be assigned different values: not what I want. I am sure Stata can do this ......

Grateful for help

Jamie Hamilton
(PhD student)