Categorical variables with large number of levels, and statistical methods for working with such variables (example: fused lasso).

Categorical variables with large number of levels, and statistical methods working with such variables (example: fused lasso). Variables with very many levels creates other problems than variables with few levels. The set of possible levels might be badly defined, so levels unseen under training might arise when using the model. Or one might want some collapsing/clustering of levels. Also use this tag for data visualization with many levels.