Currently I am doing a research about elimination of etfs. Now I have to run a logistic regression using stata. However, I read somewhere that the coefficients you get as stata output are the statistical outputs. But i guess i also need the economic significance. I know how to create those economic significance coefficients:
mfx, varlist(variables)
My question is: What is economic significance? Why do i need to calculate those when performing an logistic regression? And how should i interpret the outcomes of the mfx-code?
I hope you can help me, waiting for your comments.
Regards,
S.
The main assumption of the logistic regression is: p=Pr(y=1|x)=L(bX) where L(.) is the logistic distribution and b is the vector of parameters you are interested into. Statistical packages give you p/(1-p)=exp(bX). Suppose the j-th regressor increases by 1, then exp(bX) increases to exp(bX+bj)=exp(bX)*exp(bj) so that p/(1-p) has increased by a multiple exp(bj). For economists however, it is more natural to consider b as a semi-elasticity: ln[p/(1-p)]=bX. In practice: a logit model slope parameter of 0.1 is interpreted economically as meaning that a one-unit increase in the regressor increases p/(1-p) by a multiple 0.1. I suggest you to study these topics (and check this answer) in Cameron, Trivedi "Microeconometrics".
Hope I'v been helpful XX
So I always need to calculate the marginal effects of the logistic regression? And those marginal effects have to be multiplied by the standard deviation?
Reply With Quote
Posting Permissions
