Normality question

#1
Hi everyone. I am doing a panel data project. One of the key factors is unemployment. The dependent is program enrollment. Unemployment is not normally distributed when I look at the normality plot. Does this make regression unreliable? There is a visible pattern when I look at the plot. It's also a very large sample with a really high F.

https://goo.gl/photos/6WJLxFWL2pk8qVmc9
 
#3
Hi everyone. I am doing a panel data project. One of the key factors is unemployment. The dependent is program enrollment. Unemployment is not normally distributed when I look at the normality plot. Does this make regression unreliable? There is a visible pattern when I look at the plot. It's also a very large sample with a really high F.

https://goo.gl/photos/6WJLxFWL2pk8qVmc9
The assumption of normality in linear regression applies only to the random error term (aside from ruling out perfect collinearity, there are no real assumptions regarding the independent variables in linear regression). If this plot is for an independent variable, it's inconsequential. (If this were a plot of the residuals (estimated errors), this still wouldn't be a problem as the departure from normality is minor.)
 

hlsmith

Omega Contributor
#5
I wiil support ondansetron comment. Residuals need to be normally distributed for linaer regression modeling and time series. Though small departures will not impact the estimates just the SEs, I believe.
 
#6
I wiil support ondansetron comment. Residuals need to be normally distributed for linaer regression modeling and time series. Though small departures will not impact the estimates just the SEs, I believe.
Overall, regression is considered to be fairly robust with respect to mild-moderate departures from the assumption of normality.