Hi Team,
I ran into something that has me a little concerned. When I look, the people in the Task 2 testing population are also in the training population. I.e. there are people with the same person_id, hospitalization_date, covid_index in the task_2_goldstandard and task_2_testing_blinded_gold_standard
Is this expected? My understanding is that the testing datasets would be blinded so that we did not know the outcome. It would be very easy to 'predict' the outcome of the people in the task 2 testing datasets by looking in the training datasets.
I suspect I am looking at something wrong, but we are close enough to the deadline that I would like to discuss it with the broader community. Thanks, and I apologize if this is a waste of time!
Here is the code that is leading me to this conclusion. A single example is person_id 1012826248971412154
SELECT *
from
/*Stacked Testing and Training Datasets*/
(select person_id, hospitalization_date, covid_index, Outcome
from task_2_goldstandard
union all
select person_id, hospitalization_date, covid_index, NULL as Outcome
from Task_2_Testing_Blinded_Gold_Standard) as a
where person_id in
(select task_2_goldstandard.person_id
from Task_2_Testing_Blinded_Gold_Standard inner join task_2_goldstandard
on Task_2_Testing_Blinded_Gold_Standard.person_id=task_2_goldstandard.person_id)
order by person_id
Created by Steven Erly Serly Nevermind! I learned that the testing data will be replaced for the evaluation. Thanks!
Drop files to upload
Task 2 Blinding: I am a little concerned page is loading…