# Suggestions for Midpoint Benchmark Submissions ## Baseline Covariates vs Time-Varying Covariates **1. Features with the prefix 'BASELINE_' should not be used as Time-Varying covariates - this will cause the model to fail** **2. Features for co-morbidities (e.g. CKD) should be used as a Baseline covariate** **3. STEROIDS and DEATH must be assigned to the Binary category** ## DataTypes ** 1. Categorical features cannot be assigned to the Binary category** ** 2. Features that are numeric should be assigned to the Continuous category** ## If you have submitted a model that deviates from these suggestions, please re-submit an updated model that conforms to this guidance.

Created by Erick Scott scottie
Dear Organizers, Our team has resubmitted our models `TeamLucMP1 ` and `TeamLucMP2` last week. Could you please kindly let us know if there is any updates? Best, Haoze

Hello Tsai-Min, The following variables are time-varying: DEATH DISCHARGE_EVENT MISSING_RESP_STATUS REMDESIVIR RESPIRATORY_STATUS STEROIDS STEROIDS_CENSOR WITHDRAWAL_EVENT This information will be added to the Challenge wiki as well.
Dear organizers, As you mentioned below, could you please advice us how we can distinguish any variable that is Baseline covariate or Time-Varying covariate without seeing the detailed data point of SDY1760 dataset? Alternately, could you please distinguish the type of covariate of each feature in SDY1760 dataset for us? ``` I am sorry to report that both models could not be scored. A variable was assigned as a Baseline covariate that was not constant during the observation period. Any variable that is likely to change, i.e. medications, during the follow-up period should not be assigned as a Baseline covariate. Please re-submit the model with the dynamic variable assigned as a Time-Varying covariate. Thank you for participating. ``` Yours, Tsai-Min
Hello Haoze, TeamLucMP_1 used BASELINE_ features as time-varying. TeamLucMP_2 used features that are not varying and/or are absorbing states as Time-Varying covariates. Please review the Challenge wiki for tips on appropriate variable type assignment. Please re-submit models that fix these errors. Thank you for participating.
Dear Organizers, Thank you very much for your efforts in organizing this challenge. My team, `Team Luc`, submitted three models; however, we noticed that two of them (`TeamLucMP_1` and `TeamLucMP_2`) are missing from the results table. Could you please look into this for us? Thank you for your time and assistance. Best regards, Haoze
Hello, I am sorry to report that both models could not be scored. A variable was assigned as a Baseline covariate that was not constant during the observation period. Any variable that is likely to change, i.e. medications, during the follow-up period should not be assigned as a Baseline covariate. Please re-submit the model with the dynamic variable assigned as a Time-Varying covariate. Thank you for participating.
Dear organizers, We have submitted 2 new causal models yesterday: ``` "sdy1760_modelS" and "sdy1760_modelC" ``` , but we could not confirm whether they could be scored successfully or not. Could you please provide the link to confirm our submissions and their respective scores? Yours, Tsai-Min
Hello, The mid-point evaluation leaderboard will be released next week. Please ensure all of your models utilize the appropriate baseline or time-varying category for each feature. If a feature is likely to vary over the course of the observation period then it should be used as a time-varying covariate. Also, please ensure your submission does not contain variables from the SDY1662 dataset. Thank you.
Dear organizers, We have submitted 3 causal models: ``` "sdy1760_modelTutorial", "sdy1760_modelS" and "sdy1760_modelC" ``` , but we could not find the mid-point evaluation leaderboard. Could you please provide the link to confirm our submissions and their respective scores? Yours, Tsai-Min

Midpoint Benchmark Submission Guidance page is loading…