topic Re: The appropriate econometric model when the dependent variable is percentage (from 0 to100) in Statistical Procedures

The appropriate econometric model when the dependent variable is percentage (from 0 to100)

Aleksandar — Sat, 05 Sep 2020 10:05:12 GMT

Dear Sas community,

I was running a GLM model wherein the dependent variable is the discrete (Inv_Stock) - a percentage that takes from 0 to 100. The code is the following one:

proc glm data=WORK.test;
model Inv_Stock = Low_Target IR_minus2 IR_2 IR_4 IR_6 IR_8 IR_10 player_Target Risk_tolerance player_Posess_account player_Own_stocks Age Male Student_yes;
run;

Could you tell me whether GLM is an appropriate model for this analysis, and if not, what is the right model to use to run the regression?

Thank you,

Aleksandar

Re: The appropriate econometric model when the dependent variable is percentage (from 0 to100)

Ksharp — Sat, 05 Sep 2020 11:44:41 GMT

Check this Possion Regression.
http://support.sas.com/kb/24/188.html

Re: The appropriate econometric model when the dependent variable is percentage (from 0 to100)

sld — Mon, 07 Sep 2020 02:58:36 GMT

If the data are proportions, bounded between 0 and 1, I would consider either a binomial distribution (if the proportions are determined as a ratio of two counts) or a beta distribution. Consider the GLIMMIX procedure.

Re: The appropriate econometric model when the dependent variable is percentage (from 0 to100)

StatDave — Tue, 08 Sep 2020 13:45:25 GMT

See these notes concerning modeling continuous proportions: 56992 57480 .