Hi, How many samples are used in the bootstrapped samples? Thanks, Yang

Created by mike penc tryitout
Hi @adamklie, Bootstrap sampling is done with replacement. Does that answer your question? Cheers, Mike
Hi @domanskyi, OK, I see why there is confusion now. We should have been a bit clearer. Metric estimates are **not** computed by rerunning dockerized algorithms on bootstrapped data but by taking bootstrapped samples from participant prediction files. Hope that makes more sense, Mike
Hi @Michael.Mason, Since every bootstrap sample has repeated patients, do you make their identifiers unique in the clinical and expression data? (e.g. pat1, pat1, pat3 or pat1_1, pat1_2, pat3) Thank you, Sergii
Hi @Michael.Mason , So bootstrapping was done by taking N samples from N patients? I'm not sure I understand how you can bootstrap N from N. Could you clarify this for me? Best, Adam
@piermaro, Yes, they were made by setting the randomizing seed to the same number at the beginning of each procedure. Regards, Mike
Dear @domanskyi and @tryitout Bootstrap sampling was done by taking N sample from N patients with replacement **ten times**. Where N = 138 and 137 for the nivo and chemo arms respectively. Hope this helps, Mike
Dear @vchung , I have a related clarification question: were the results for the Baseline Models posted in the Leaderboard obtained with exactly the same "10 bootstrapped samples" procedure? Thanks! Carlo
Hi, Does it mean that you randomly select 10 patients (without replacement) and then the model is ran and performance metrics calculated? Then new 10 patients are randomly re-selected and this is repeated N times? How big is N? The average score is then assigned to the submission? Thank you,
Hi @vchung , Is that mean only 10 patients are used for each bootstrap evaluation set? My understanding of 10 bootstrapped samples is that it was resampled 10 times from all the patients and my question is how many patients are resampled each time? Thanks, Yang
Hi @tryitout , Thank you for participating in the Challenge! The returned estimates are based on 10 bootstrapped samples. More information can be found under [The Challenge > Assessment](https://www.synapse.org/#!Synapse:syn18404605/wiki/607226).

How big are the bootstrapped samples? page is loading…