Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# Re: st: Calculating and interpreting effect size when DV is a proportion

 From Maarten Buis To statalist@hsphsun2.harvard.edu Subject Re: st: Calculating and interpreting effect size when DV is a proportion Date Mon, 14 Jan 2013 09:42:27 +0100

```--- On Sat, Jan 12, 2013 at 7:50 PM, Michelle Dynes
>> I am conducting analyses using proportion data as my outcome. I am
>> using GLM with a logit link and the binomial family to carry out the
>> analyses. I would like to determine the effect size of my independent
>> variables. Here are two options I have considered:
>>
>> Generate Odds Ratios using the -eform- command at the end of the GLM
>> statement. Does this make sense to do when my outcome is not
>> dichotomous?

On Mon, Jan 14, 2013 at 7:41 AM, Jeffrey Wooldridge wrote:
> I would use the "margins" command to compute the average marginal
> effects. That we you do not have to fix the explanatory variables
> (continuous or binary) at any specific values. For the binary
> variables, you get the change in the predicted y when the dichotomous
> variable goes from zero to one. For the continous variables you get
> the effect of a one-unit increase in x (provided that is a reasonable
> change and the calculus approximation is sensible).

You can also use the exponentiated coefficients, but they are not odds
ratios. In the past I have called these "relative proportion ratios".
The interpretation of these numbers is exactly the same as in
-betafit- and in the helpfile of -betafit- there is a discussion of
these coefficients. You can install -betafit- (and thus obtain its
helpfile) by typing in Stata: -ssc install betafit-. I have also
applied this logic to a fractional logit model in this post on
Statalist: <http://www.stata.com/statalist/archive/2013-01/msg00011.html>.

Hope this helps,
Maarten

Ps. I hope Michelle has also included the -vce(robust)- option to her
-glm- command. This is a crucial part of the fractional logit model.

---------------------------------
Maarten L. Buis
WZB
Reichpietschufer 50
10785 Berlin
Germany

http://www.maartenbuis.nl
---------------------------------
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/
```