search

wustat / rjaf

Regularized Joint Assignment Forest
0 stars 0 forks source link

Issues when testing the function with real data #4

Open rladhania opened 2 months ago

rladhania commented 2 months ago

Issues that popped up while testing (will keep updating this list as more come up)

  1. when clus.tree.growing = T, the function is executed only when the treatment arm is named as "trt". Any other name such as "arm", even when explicitly specified in the function call (trt = "arm") throws up an error. This problem doesn't arise during the unclustered execution (clus.tree.growing = F).
  2. When clus.tree.growing = T, can we also send as function output the treatment clusters (which arms were clustered together)? That will be useful info for the social science researcher/application folks to get a sense of which arms are being clustered together, and if any meaningful patterns arise.
  3. Can we make the prob argument optional?
XinyiEmilyZhang commented 1 month ago

@rladhania Thanks for checking! I’ve resolved the potential issue related to Issue 1 and tested it with sample data where "arm" was used as the treatment name. Could you confirm if this fix works with the real dataset as well?

For Issue 2, if you were referring to obtaining treatment clusters based on the rjaf algorithm, it could be obtained by the column in output dataset called "trt.rjaf". Or, if you want to obtain the treatment cluster from the original treatment, you can merge the output dataset with the original data by id. Let me know if anything is unclear.