In an AI model, what is the proper way to normalize by industry ?
If I use something like:
ZScore("formula“, #industry, 7.5, NA, 5)
Will it be properly normalized or will the model use the zscore of the zscore
Currently we can only normalize by sub-sector, why ?
You can normalise by Industry in the way you just mentioned but I think you’d then want to skip normalisation in the dropdowns to avoid taking the z score of the z score (which might produce the same values anyway but could get messy with the outlier/trim settings). Can also try FRank(“formula”,#Industry,#ASC/#DESC) if magnitude isn’t a concern.
That said, the tricky thing about using #Industry, and maybe the reason it isn’t an option on the dropdown, is because there are industries with not many stocks in it. I think this would really mess up the distributions for the z scoring.
Thanks, makes sense
Yes we skipped for the reason that @trendyist said, but was not tested. Maybe it's great? We should add it for completeness; pretty easy to do. The workaround is to click on Add Formula then choose 'Macro Ratio' in Formula Type. It's strange, but that's the only way right now to skip both normalization steps.
Thanks, one more question, when we normalize by industry or subsector, the normalization includes only stocks that are in the model Universe right ? Is there a way to use all stocks in the same sub-sector/industry ?
