if om was short for only_mean then we have to update this to oc for only_central

function name summarise and summarize looks confusing. can't we suggest a better name for this function? If we don't want to interfere with how this is used outside utils we can rename summarise?

also, do we really need this function? I think this is just calling/copying summarise function with an argument mean instead of the default median. Can't we have summarise default to mean on central_measure and delete this function? in that case summarise will be renamed to summarize

I agree with @mnjowe that having two different but similar functions that differ only the in whether they use an ize or ise is likely to be confusing.

I suspect the reason you've kept summarize with the original default behaviour of computing mean and same call signature as previously @tbhallett is to ensure this doesn't break code using the previous version?

There are few alternative ways we could deal with this:

• Just making a breaking change and removing the old function and using summarize for new function (potentially with a helpful error message if users try to call with previous signature).
• Using a more differentiating name for new function - perhaps something like compute_summary_statistics?
• Changing default behaviour of new function to replicate old behaviour and using same summarize name, but raising deprecation warning if relying on old defaults. This would require doing somethin like defaulting tocentral_measure="mean", adding a **kwargs argument to new function, checking if only_mean is in kwargs and using to set value of only_central as well as raising a deprecation warning to indicate that this argument is deprecated and users should use use_central instead, and also dealing with the adjustment to column name from central to mean in this case. Overall this is probably a bit complex.

can't we do this in summarise function, delete this function and rename summarise to summarize?

passing only_mean to only_central here looks confusing. can't we use one name for consistency?

Small update to fix parameter directive syntax in docstring and adding return information.

Indexed assignment to a dict is generally preferable over update when just adding a key-value pair as it avoids creating an unecessary intermediate dictionary and is slightly more readable.

We could also avoid the repetition across the different conditions by changing to something like

    if central_measure in ('mean', 'median'):
        grouped_results = results.groupby(axis=1, by='draw', sort=False)
        stats['central'] = getattr(grouped_results, central_measure)()

but I think on balance probably the loss of readability outweighs the slight gain in avoiding code redundancy.

To avoid computing the groupby operation multiple times here and when computing central measure, it would be better to have one line something like

grouped_results = results.groupby(axis=1, by='draw', sort=False)

at beginning of function and then using grouped_results in place of repeated results.groupby(axis=1, by='draw', sort=False) calls.

I would possibly also say writing this as two separate indexed assignments to the stats dict rather than using update would be a bit more readable

lower_quantile = (1 - width_of_range) / 2
stats["lower"] = grouped_results.quantile(lower_quantile)
stats["upper"] = grouped_results.quantile(1 - lower_quantile)

but using dict.update here to update dictionary with two entries is reasonable so I think this is more of a personal preference thing!

I agree with @mnjowe that having two different but similar functions that differ only the in whether they use an ize or ise is likely to be confusing.

I suspect the reason you've kept summarize with the original default behaviour of computing mean and same call signature as previously @tbhallett is to ensure this doesn't break code using the previous version?

There are few alternative ways we could deal with this:

• Just making a breaking change and removing the old function and using summarize for new function (potentially with a helpful error message if users try to call with previous signature).
• Using a more differentiating name for new function - perhaps something like compute_summary_statistics?
• Changing default behaviour of new function to replicate old behaviour and using same summarize name, but raising deprecation warning if relying on old defaults. This would require doing somethin like defaulting tocentral_measure="mean", adding a **kwargs argument to new function, checking if only_mean is in kwargs and using to set value of only_central as well as raising a deprecation warning to indicate that this argument is deprecated and users should use use_central instead, and also dealing with the adjustment to column name from central to mean in this case. Overall this is probably a bit complex.