Regressions with two independent variables where one is part of another.

33 Views Asked by At

I am doing simple regression analysis where I use one-way fixed effect model to estimate the effects of two variables on dependent variable. The question I am asking is how to interpret these two variable's coefficients after regression if one is subset of another, e.g A is a part of B.

For example, I want to research how the growth of yearly purchases affected investment policy. And I am able to decompose the purchases on different types of purchases. Lets assume there are 10 types of purchases and I am interested particularly in type D purchases.

How do I interpret result if for example I would the regress investments policy on total purchases and D purchases growths, and then only D purchases growth.

I was told that if you have for D purchases growth coefficient (-/minus) then the effect of D purchases is less effective than total purchases; (+/plus) then it increases the effect of total purchases; (insignificant) then there is no effect outside total purchases.

Am I wrong? Is there any more convenient explanation of it? Could you help and explain, please?

Regression in Stata:

xtreg Invest L.(Invest D_purchases Leverage SizeAssets ROA TobinQ Tangible Growth), fe vce(cluster eTicker)

xtreg Invest L.(Invest Total_purchases D_purchases Leverage SizeAssets ROA TobinQ Tangible Growth), fe vce(cluster eTicker)