Regress with weighting

[Martin Imelda Borg]

Hello there

I would like to use my weights in my regression
I am not sure if it's possible, is it possible to see the values for the base category ?

I would like to know it there is a difference between high-performance athletes and low performance athletes in terms of nutrition (categorical variable), age (conitnous) gender (categorical) , training location (3 level categorical)
categorical variable = 0 (low performance), 1 (high performance)

reg performance age i.gender i.location i.nutrition [pw=attweight]

As expected the results are only given for my high performance athletes = 1 (I assume....) This means that if coef for age is -0.001 p value 0.87. This means that high performance athletes are slightly younger but there is no significant difference. Is this correct?

1. How do I obtain odds ratios - as when I try

logit i.performance age i.gender i.location i.nutrition, or [pw=attweight]

error: [  not allowed

2. As my training location are 3 levels categorical variable, I only obtain the values for 2-3. Can I get the results for 1 ?

3. Why am I not able to use

logit i.performance age i.gender i.location i.nutrition [pw=attweight]

Sata says not able to use factor variable - but the results are given when I omit the I.
Is this wrong if I omit the -i- ? Will stata consider this as a continuous variable?

4. What is the difference between -regress- and -logit-?
As in both they produce the same similar results

[Daniel Feenberg]

What makes you think you only have results for performance==1? If you run -predict- you will see predicted values for every record.

To get logit predicted values for choice 1, see https://www.stata.com/manuals/rlogitpostestimation.pdf

With only two choices on the dependent variable, -logit- and -reg- often provide similar results. They would have very different responses to extreme values on the RHS. -reg- might give predicted values less than zero or more than one. How do you feel about that? If a predicted value of 1.5 makes sense as "medium performance" then maybe you can stick with -reg-. -logit- was developed for the case of transportation choice. If bus is 0 and car is 1, 1.5 didn't make any sense as a predicted value.