heteroscedastcity

noetsi

Fortran must die
#1
1562005034990.png

My residual plots suggest heteroscedastcity to me (fisher should have lost his knighthood for that horrible word). My question is, is it enough to use white's standard errors to deal with this problem. I have about 5 thousand cases if that matters.
 

Dason

Ambassador to the humans
#2
I'm not actually convinced there is any issue here. Maybe a few outliers on the left but I don't think heteroscedasticity is really an issue. When you have more observations in an area one expects to see some residuals get to higher values. It looks like you have more observations with low days in the program and less with high days in the program.
 

noetsi

Fortran must die
#3
thanks dason. I figured since the range of the data (to me anyhow) seems to be narrowing there is a problem. I have read about a 'fan shape' reflecting problems and figured this was that :p

Its hard for me to go from the ideal types I see in text to what I actually find in my data.