## Strange Regression Results

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

KielanHelp
Posts: 1
Joined: Sun Feb 25, 2018 6:22 am

### Strange Regression Results

Hey Everyone! This is my first time writing and I'm an absolute beginner so please explain thoroughly.

I am running a regression on a child labour survey. I have chosen specific questions and turned them into a binary code. For example, a question might be Highest level of education achieved by parent, and I have split this up into None(1,0) , Upper (1,0) , or Lower (1,0). I did this for a couple of questions. This makes them all dummy variables. When I eventually put all my variables into a regression, to determine the dependent variable of school attendance(1,0), it makes some coefficients negative that should be positive such as a being in a wealthy family. Here is a picture.

I used series NAME = @recode(Work = 1,1,0) to code all of this

I can include my full regression through eviews if anyone wants to help!

Just the wage regression is different from the full regression, as one coefficient is positive and the other negative
Attachments
Wage gap.PNG (17.66 KiB) Viewed 434 times
Full Regression.PNG (36.82 KiB) Viewed 434 times