[Python] Pandas time series dataframe take random samples by group (date) ignore missing dates

I have a time series dataframe, I would like to take x random samples from column "temperature" from each day. I am able to do this with:

daily_groups = df.groupby([pd.Grouper(key='time', freq='D')])['temperature'].apply(lambda x: x.sample(10))


This works if there are at least x samples for each day. If there are not I get the error "a must be greater than 0 unless no samples are taken". What I would like is, lets say I want 10 samples, get 10 if available, if not get as many as possible and if there are none skip this day. I don't want to up sample data.

Also I don't know if its possible to return the original dataframe with the values filtered for items mention above, instead of returning the groupby series object. Thanks for any help.

Continue reading...