Analysis of Conservative Effects: research code repository for developing a new data analysis method for controlled experiments. Results can be classified as practically significant, practically insignificant, or inconclusive.
0
stars
0
forks
source link
Combinatorics speed up for statistic calculations in agreement contest #7
One way to massively speed up the candidate statistic calculations from the simulated samples would be only to compute a small fraction of samples for the experiment and control group, and then compare the values across a large number of pairwise combinations. The challenge is the groundtruth would have to be computed for all of these combinations, which would be complicated to implement.
We would have to add a decision rule of each measure of disagreement (a function?) and record which columns to compare? And then grab values from the columns in combinations and then run them through the function.
One way to massively speed up the candidate statistic calculations from the simulated samples would be only to compute a small fraction of samples for the experiment and control group, and then compare the values across a large number of pairwise combinations. The challenge is the groundtruth would have to be computed for all of these combinations, which would be complicated to implement.
We would have to add a decision rule of each measure of disagreement (a function?) and record which columns to compare? And then grab values from the columns in combinations and then run them through the function.