How to regress categorical variables without base

[willemijn gr]

Hello,

Currently I am investigating the impact of industry categorization (categorical variable) on premium's paid (continuous variable) in acquisitions. Therefore, I have categorized my data in 48 industry categories and now I am trying to regress the industries with the paid premiums. But stata automatically selects one of my industries as base. How can I calculate the impact of the different industries on the premiums without a base since there is no standard industry?

Thank you for your help!

[Rich Goldstein]

you don't show your code so I can't show you what to change; in general, however, where you currently have i.varname switch to ibn.varname; for the meaning of this, see

help fvvarlist

and look at the section on selecting the base level

[Eric de Souza]

"How can I calculate the impact of the different industries on the premiums without a base since there is no standard industry?"
You cannot unless you drop the intercept (constant term). One sector will always be excluded but you can choose which one as Rich Goldstein indicate. Search for "dummy variable trap" either in your browser or in a textbook.

[Carlo Lazzaro]

Willemijn:
welcome to this forum.
Consistently with Eric's helpful advice, see https://en.wikipedia.org/wiki/Dummy_...le_(statistics).

[ericmelse]

Dear Willemijn,

To compare and analyse all categories, including the base or reference category, you could apply the so-called quasi-variance method, proposed by David Firth, see for his personal website here.
His implementation is in R, but a Stata package is available, developed by Aspen Chen:

net describe qv, from(http://fmwww.bc.edu/RePEc/bocode/q)

For background reading you can read these papers:
Social Methodology. 33 (1): 1-18, D. Firth (2003). Overcoming the reference category problem in the presentation of statistical models. The working paper version you can find here.
Biometrika. 91 (1): 65-80, D. Firth & R.X. de Menezes (2004). Quasi variances.

There is Wikipedia lemma available here.

Some more Stata and SPSS resources are available here, made available by Vernon Gayle and Paul Lambert who published the following (must read) paper:
Sociology. 41 (6): 1191–1208, V. Gayle & P. S. Lambert (2007). Using Quasi-variance to Communicate Sociological Results from Statistical Models. The working paper version you can find here.

Best,
Eric

[Marcos Almeida]

You could just use margins and marginsplot.