Problem with residual?

noetsi

Fortran must die
#1
I don't know if this is an issue or not. There are the strange bars on the top and bottom of the residuals - I have never seen anything like this and can't imagine substantively what would cause it. My DV here is hours worked which is not going to be more than 40 or less than 0 normally.
 

noetsi

Fortran must die
#4
My out come is whole hours. No customer has more than 40 hours and very few have negative hours. There are a wide range of predictors some interval and some dummy variables. I am guessing this is caused by range restrictions on the DV.
 

noetsi

Fortran must die
#5
This isn't unusual to see when the DV is discrete.
It has about 40 different levels.

I guess my next question, since I have not seen this addressed, is this a serious issue for the standard errors. I have about 11000 data points if that matters....
 

hlsmith

Omega Contributor
#6
You are probably fine like normal, but is there a slight negative trend in these residuals? So they are not some IID.


Well I always thought concerns now come in regards to how you interpret coefficients that say with variable X, y-hat increases 4.65434. Are you fine with decimals in the face of using integers. Most times it is just a round procedure needed - I would guess?
 

noetsi

Fortran must die
#7
That is a good point about interpretation of the slopes when you have data that can't be subdivided. Is it reasonable to say an increase in 4.6 hours when in fact you can only work whole hours? I actually never considered that before... I honestly don't know if it is legitimate to talk about the data that way. Or whether rounding solves the problem.

Sometimes it is painful to come to this board :p Makes me think to much.