The key issues are probably these:
- Are there separate support variables used anywhere? If so, have they been the original integer in the range distributions of 0-6 or somehow modified? Eg. Categorized?
- Is the "sum of the amounts" used as such, converted to one of the variable conversion or categorized?
The problem is that the distribution in terms of individual donors is a difficult shape: zero and six is often the case where there is a big spike. Distributions on the neighbor or the "other" obtained are such that zero is a big spike and the other values are very rare. Can those assigned to groups? Do you have experience?
Thank you, Minna Salakari