please help : modelization

#1
Hello,

I have a distribution which looks like a Khi-2 distribution, which model is most appropriate?

I try linear regression, but normal distribution is not suitable.

Thanks for your help ! :)
 
#4
So, the distribution more looks like a normal distribution, but the error of linear model increase, I will continue to search but if you have other idea I'll take it.
 

hlsmith

Omega Contributor
#5
Well it doesn't matter too much if the y-variable itself is normally distributed. You just need the assumptions about the model error term to hold. So you should be working to normalize them and get homoscedasticity!