I’m currently applying the for loop below to a pandas dataframe with fields prob, id, and time. How can I vectorize this?

for cutoff in np.sort(df['prob'].unique()):
    # get subset of records where prob >= cutoff
    sub = df[df['prob'] >= cutoff]
    # from that subset, for each ID get the record with minimum prob 
    subsub = sub.loc[sub.groupby('id')['prob'].idxmin()]
    # for those records, compute various statistics on the time field
    times = subsub['time'].values
    data.append([cutoff, np.quantile(times, 0.5), np.quantile(times, 0.9), (times <= 7).sum()/len(times)])
metrics = pd.DataFrame(data)