11

I have a table like this

    timestamp   avg_hr  hr_quality  avg_rr  rr_quality  activity    sleep_summary_id

    1422404668  66      229             0       0           13              78
    1422404670  64      223             0       0           20              78
    1422404672  64      216             0       0           11              78
    1422404674  66      198             0       40          9               78
    1422404676  65      184             0       30          3               78
    1422404678  64      173             0       10          17              78
    1422404680  66      199             0       20          118             78

I'm trying to group the data by timestamp,sleep id and rr_quality, where rr_quality is > 0

I've tried the following and none of them seems to work

 df3 = df2.groupby([df2.index.hour,'sleep_summary_id',df2['rr_quality']>0])

 df3 = df2.groupby([df2.index.hour,'sleep_summary_id','rr_quality'>0])

 df3 = df2.groupby([df2.index.hour,'sleep_summary_id',['rr_quality']>0])

All of them returns a keyerror.

EDIT:

Also can't seem to be able to pass more than one filter at a time. I tried the following: 

df2[df2['rr_quality'] >= 150, df2['hr_quality'] > 200]
df2[df2['rr_quality'] >= 150, ['hr_quality'] > 200]
df2[[df2['rr_quality'] >= 150, ['hr_quality'] > 200]]

returns: TypeError: 'Series' objects are mutable, thus they cannot be hashed

cyberbemon
  • 3,000
  • 11
  • 37
  • 62

2 Answers2

11

the simplest thing to do here is to filter the df first and then perform the groupby:

df2[df2['rr_quality'] > 0].groupby([df2.index.hour,'sleep_summary_id'])

EDIT

If you're intending to assign this back to your original df:

df2.loc[df2['rr_quality'] > 0, 'AVG_HR'] = df2[df2['rr_quality'] >= 150].groupby([df2.index.hour,'emfit_sleep_summary_id'])['avg_hr'].transform('mea‌​n')

The loc call will mask the lhs so that the result of the transform aligns correctly

To filter using multiple conditions you need to use the array comparision operators &, | and ~ for and, or and not respectively, additionally you need to wrap the conditions in parentheses due to operator precedence:

df2[(df2['rr_quality'] >= 150) & (df2['hr_quality'] > 200)]
EdChum
  • 376,765
  • 198
  • 813
  • 562
2

I know this is old but I wanted to add that there is an official function to do exactly this. Transforming the example from pandas to your case:

grouped_df2= df2.groupby([df2.index.hour,'sleep_summary_id','rr_quality'])
grouped_df2.filter(lambda x: x['rr_quality'] > 0.)
Czarking
  • 143
  • 1
  • 10