[BUG] rake doesn't support trimming - but also doesn't indicate it to the user

[talgalili]

This is violating user expectation.

E.g.:

weights_untrimmed = sample.adjust(
    variables=weighting_variables,
    method="rake",
    weight_trimming_mean_ratio=0,
    transformations=auto_recodes,
)

By expectation violation I mean,

1. there isn’t anything in the docs that suggests that rake doesn’t support weight_trimming_mean_ratio or weight_trimming_percentile
2. there isn’t an error thrown when a value is passed to either weight_trimming_mean_ratio or weight_trimming_percentile suggesting that rake doesn’t support it.
3. this can result people thinking the weights are getting trimmed when they aren’t.

(Reported by David Lovis-McMahon)

[talgalili]

This issue also happens for post stratification.
The locations of the functions to fix:

balance/balance/weighting_methods/rake.py

Line 26 in 0081b51

def rake(

balance/balance/weighting_methods/poststratify.py

Line 20 in 0081b51

def poststratify(

We might need to add it here:

balance/balance/weighting_methods/rake.py

Line 243 in 0081b51

# rescale weights to sum to target_sum_weights (sum of initial target weights)

As opposed to how it's solved for, say, CBPS:

balance/balance/weighting_methods/cbps.py

Line 676 in 0081b51

weights = balance_adjustment.trim_weights(

And ipw

balance/balance/weighting_methods/ipw.py

Line 178 in 0081b51

weights = balance_adjustment.trim_weights(

Since there is always also the stage of normalizing the sum of weights of target population, it might be more easily solved as a new function that trims and normalize to sum of weights of the target.