# Ordinal logistic regression: g-o-f test in Stata13

#### layacoop

##### New Member
I'm running an analysis on ordinal data in Stata13. I'm looking at 7 different DVs (separately), so parsimony is crucial--I will be repeating every analysis 7 times.

For each DV, I have 9 IVs (ordinal, count, and binary categorical).

I'm having 2 main problems:

1) I created composite scores for 3 variables (otherwise I would have upward of 20 DVs). However, after using the --tab varname -- command I think I may have done this incorrectly (for some items it gives me upward of 50 levels). What I want is for 3 items each on a 4-point scale to be turned into 1 item on a 12-point scale. Does anyone know how to do this? (I ran a Chronbach's alpha to make sure I can combine the items into one term).

2) I'm having trouble running a g-o-f test for my model. It seems Chi-square is the best thing to use here, but the command (both estat gof and chi2) doesn't seem to be working. Is this the wrong command for this version of stata? Also, for ologit, does a gof est come just at the end, or after each step in a backward elimination process?

Any help i greatly appreciated!
Thanks!

#### Phaedrus

##### New Member
Regarding your first question: after typing alpha in Stata you can generate a new variable with the generated scale. So it would be:

Code:
 alpha x1 x2 x3, gen(newname)