Statistical test for ordinal variable distribution between groups I have a dataset with sentiment of tweets labelled as 1, 2 or 3 (1=Negative, 2=Neutral, 3=Positive) ... 10,000 observations approx.
I have this split by geographic location (Africa, Americas, Asia, Europe, Oceania.)
I was looking for an appropriate statistical test to compare sentiment between the geographic groups.
I had thought Friedman but this requires that I have the same number of observations in each group.
Can anyone point me to an appropriate statistical test to compare the samples ?
Edit:
geo       label
Africa    1           6
          2          27
          3         297
Americas  1         356
          2         809
          3        3353
Asia      1           3
          2         171
          3         345
Europe    1         147
          2         479
          3        2518
Oceania   1          60
          2         372
          3        1063

 A: You could look into ordinal logistic regression, but first you shoud do some visualization. A mosaicplot of the table is

But some formatted tables of percentages by row can be useful. Here is the contingency table:
          sentiment
region        1    2    3
  Africa      6   27  297
  Americas  356  809 3353
  Asia        3  171  345
  Europe    147  479 2518
  Oceania    60  372 1063

and here with row percentages:
          sentiment
region     1  2  3
  Africa   2  8 90
  Americas 8 18 74
  Asia     1 33 66
  Europe   5 15 80
  Oceania  4 25 71

and given the large sample, it is quite clear that the profile of sentiment is unequal among regions. We can get a likelihood ratio test via ordinal logistic regression in R:
mod.polr <-  MASS::polr(label  ~  geo, data=byregion, weights=count, Hess=TRUE)
mod.null <-  MASS::polr(label  ~  1, data=byregion, weights=count, Hess=TRUE)
anova(mod.null, mod.polr)   

Likelihood ratio tests of ordinal regression models

Response: label
  Model Resid. df Resid. Dev   Test    Df LR stat. Pr(Chi)
1     1     10004   13745.75                              
2   geo     10000   13632.95 1 vs 2     4 112.8001       0

