0

In typical ANOVA, we have deviation of group means from grand means (sum of squared treatment). If data is unbalanced, would it still be the deviation of group mean from grand mean (weighted by number of observations)? Or, would it instead be the deviation of group mean from the average of the group means?

Is there a good resource that goes into talking about the intuition and math behind unbalanced groups? I have seen posts saying analysis is fine as long as groups are normal and have common variance.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
confused
  • 2,453
  • 6
  • 26
  • 1
    Many similar Qs, for instance https://stats.stackexchange.com/questions/122974/are-unequal-groups-a-problem-for-one-way-anova and [this list](https://stats.stackexchange.com/search?q=one-way+anova+unbala*+answers%3A1). – kjetil b halvorsen Sep 15 '20 at 14:27

1 Answers1

0

This seems to have most of the answers: One way unbalanced Anova and the sum to zero constraint

The formula is Total Sum of Squares = Treatment Sum of Squares + Error Sum of Squares. Seems like no matter how you specify the model, Treatment Sum Of Squares stays the same.

confused
  • 2,453
  • 6
  • 26