3

I have attached a residual plot (and QQplot), and i would like to know if you think its too big of a problem? It is not fully equally spread around 0 and there is a fairly big outlier.

all the best

enter image description hereenter image description here

enter image description here

Ertxiem - reinstate Monica
  • 1,403
  • 4
  • 8
  • 17
maS
  • 139
  • 1
  • 7
  • Would you please add a histogram plot of the residual error? – James Phillips May 02 '19 at 09:28
  • 1
    Hi James. Sorry, i dont know how to upload an additional image. However you can see a histogram of the residuals on this link: https://ibb.co/qR1TsRq – maS May 02 '19 at 09:41
  • 1
    I personally think this is not a bad fit. It could be better, but I have certainly seen worse. – James Phillips May 02 '19 at 15:00
  • 1
    Possible duplicate of [Interpreting QQplot - Is there any rule of thumb to decide for non-normality?](https://stats.stackexchange.com/questions/111010/interpreting-qqplot-is-there-any-rule-of-thumb-to-decide-for-non-normality) – mkt May 03 '19 at 13:44
  • 3
    I have upvoted this question because the low outlying fitted value is a concern that I haven't seen addressed in other questions about interpreting residual plots. (cc @mkt). Furthermore, although there should be little concern about non-Normality affecting the computation of p-values or confidence limits, the presence of seven unusually high residuals is worth a separate investigation not because they indicate non-Normality but because they appear to be systematically different from the rest of the data and it might be informative to know why. – whuber May 03 '19 at 14:14
  • 2
    My guess is that https://www.statalist.org/forums/forum/general-stata-discussion/general/1494766-heteroscedasticity-vs-homoscedasticity-with-different-measures-of-dependent-variable provides context. Further guesses are that the OP has totals of e.g. deaths for one or more regions and is not scaling them for size of population at risk. Working on log scale (?) is dampening variability but not correcting for this. Whether my guesses are right or wrong, I agree with @whuber that there is structure here and information about what underlies the plots is essential for good advice to be possible. – Nick Cox May 03 '19 at 14:52
  • I think it's 8 big residuals. @maS in Stata specify e.g. `ms(Oh)` so that the overlap and occlusion of data points is reduced. – Nick Cox May 03 '19 at 14:53

0 Answers0