Suggestion: percentile ranks

[MichaelWS]

I realize I am computing percentile ranks constantly in my code.

df["pct_rank"] = df["field"].groupby("date").transform(lambda x: x.rank(ascending=False) / float(x.count()))

Would anyone have any use for a function that is computed in cython for this?

if so, would people prefer to it to be a separate function or an option in rank?

[jreback]

can you do this with cumcount and size?

[MichaelWS]

Thanks. You are right.

You could do it by this
grouped = df["field"].groupby("date")
df["pct_rank"] =grouped.rank() / grouped.size()

though its a 15% speedup to put it into the rank() function.

[jreback]

true....you could add a kw arg to the groupby rank..easy enough

and get experience with cython...always good!

go for it

maybe percentage=False as a kw

[MichaelWS]

Thanks. I added that as a keyword for a series.

groupby("date").rank(percentage=True) works as intended for me.