Timeline for Handling NaN values by replacing them with -999
Current License: CC BY-SA 3.0
4 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Nov 10, 2017 at 9:52 | comment | added | Stephan Kolassa | Why shouldn't your decision trees split spuriously on this predictor, especially if missingness is possibly not at random? | |
| Nov 10, 2017 at 9:42 | comment | added | Toutsos | What i meant with "I am not planning on running a parametrical model so i suppose the -999 value will not really hurt my model." is that i will be mostly using decision trees. If i used a regression, the -999 would have a huge impact but with a decision tree, i am inclined to believe the splits wont be affected that much or they will at least be affected positively. | |
| Nov 10, 2017 at 8:28 | vote | accept | Toutsos | ||
| Nov 10, 2017 at 7:30 | history | answered | Stephan Kolassa | CC BY-SA 3.0 |