Timeline for Can I use a standard scaler to compare categorical and non categorical variables in regression
Current License: CC BY-SA 4.0
3 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 30, 2022 at 20:11 | comment | added | EdM | @Willard yes. But if you don't do that you almost always end up with problems. Unless you represent one level as all 0s you end up with a linearly dependent set of one-hot-encoded columns in the design matrix. With a simple model you might get away with that by removing the intercept to obtain estimated outcomes directly for each level. In more complicated models, software will either refuse to handle the linearly dependent predictors or remove its own choice of one of them, perhaps even silently. Safest to choose your own reference level and do comparisons of interest after modeling. | |
| Oct 30, 2022 at 18:37 | comment | added | Willard | Is it true that that a one hot encoded dummy variable represents the difference from a reference only if one col is held out (example: one category is represented by 00 in a 3 category one hot encoded transformation). | |
| Oct 27, 2022 at 16:30 | history | answered | EdM | CC BY-SA 4.0 |