Jake
Jake

Reputation: 474

Looping through a python dictionary and manipulate each value

I am a fairly new python user and I am stuck on a problem. Any guidance would be greatly appreciated.

I have a pandas data frame with three columns 'ID', 'Intervention', and 'GradeLevel'. See code below:

 data = [[100,'Long', 0], [101,'Short', 1],[102,'Medium', 2],[103,'Long', 0],[104,'Short', 1],[105,'Medium', 2]]

intervention_df = pd.DataFrame(data, columns = ['ID', 'Intervention', 'GradeLevel'])

I then created a dictionary of data frames grouped by 'Intervention'. See code below:

intervention_dict = {Intervention: dfi for Intervention, dfi in df.groupby('Intervention')}

My question is can you loop through the values of the dictionary and manipulate each value of the dictionary? Specifically I am trying to reference a look-up table. The lookup table can be thought of as a roster. My goal is to label anyone in the roster as either 'Yes - Intervention Name' or 'No Intervention'. It becomes tricky because lets say the Long Intervention, for instance, has only GradeLevel 0. That means I would want to tag anyone in the intervention_df with grade level 0 as 'Yes - Long' and anyone not in the intervention_df as 'No - Long' this would become a new column called 'Value'. I would also need to create another variable 'Category' which would specify the intervention name in this example it would simply be 'Long'

lookup_data = [[100, 0], [101, 1],[102, 2],[103, 0],[104, 1],[105, 2], [106, 0], [107, 0],[108, 2],[109, 1]]
lookup_df = pd.DataFrame(lookup_data, columns = ['ID', 'GradeLevel'])

For example the 'Long' dictionary would look like this after the processing:

longint_data = [[100,'Long', 'Yes - Long'],[103,'Long', 'Yes - Long'], [106,'Long', 'No - Long'], [107,'Long', 'No - Long']]
longint_df = pd.DataFrame(longint_data, columns = ['ID','Category', 'Value'])

The desired final output after all manipulation would look like this:

result_data = [[100,'Long', 'Yes - Long'] , [101,'Short','Yes - Short'], [102,'Medium','Yes - Medium'], [103,'Long', 'Yes - Long'], [104,'Short','Yes - Short'] , [105, 'Medium','Yes - Medium'], [106,'Long', 'No - Long'], [107,'Long', 'No - Long'], [108,'Medium','No - Medium'], [109,'Short','No - Short']]

result_df = pd.DataFrame(result_data, columns = ['ID','Category', 'Value'])

Thank you!

Upvotes: 1

Views: 83

Answers (2)

Andy L.
Andy L.

Reputation: 25269

Here the solution without using dictionary intervention_dict. Below is your data which I get from your commands:

In [1048]: intervention_df
Out[1048]:
    ID Intervention  GradeLevel
0  100         Long           0
1  101        Short           1
2  102       Medium           2
3  103         Long           0
4  104        Short           1
5  105       Medium           2

In [1049]: lookup_df
Out[1049]:
    ID  GradeLevel
0  100           0
1  101           1
2  102           2
3  103           0
4  104           1
5  105           2
6  106           0
7  107           0
8  108           2
9  109           1

Step 1: Doing outer merge between lookup_df and intervention_df, create column Value and set_index to GradeLevel

In [1059]: df = lookup_df.merge(intervention_df, on=['ID', 'GradeLevel'], how='outer').assign(Value='Yes - '+intervention_df['Intervention']).set_index('GradeLevel')

In [1060]: df
Out[1060]:
             ID Intervention         Value
GradeLevel
0           100         Long    Yes - Long
1           101        Short   Yes - Short
2           102       Medium  Yes - Medium
0           103         Long    Yes - Long
1           104        Short   Yes - Short
2           105       Medium  Yes - Medium
0           106          NaN           NaN
0           107          NaN           NaN
2           108          NaN           NaN
1           109          NaN           NaN

Step2: create df_fillna to fill NaN in df

In [1063]: df_fillna = intervention_df.groupby('Intervention').head(1).assign(Value='No - '+intervention_df['Intervention']).set_index('GradeLevel')

In [1064]: df_fillna
Out[1064]:
             ID Intervention        Value
GradeLevel
0           100         Long    No - Long
1           101        Short   No - Short
2           102       Medium  No - Medium

Step 3 (final): using combine_first to fill NaN in df from df_fillna values and reset_index to delete 'GradeLeveland doingsort_valuesonID`

In [1068]: df.combine_first(df_fillna).sort_values('ID').reset_index(drop=True)
Out[1068]:
    ID Intervention         Value
0  100         Long    Yes - Long
1  101        Short   Yes - Short
2  102       Medium  Yes - Medium
3  103         Long    Yes - Long
4  104        Short   Yes - Short
5  105       Medium  Yes - Medium
6  106         Long     No - Long
7  107         Long     No - Long
8  108       Medium   No - Medium
9  109        Short    No - Short

Upvotes: 1

Matt W.
Matt W.

Reputation: 3722

This is what I feel like you're going for.. but without more clear explanation, I"m not sure.

data = [[100,'Long', 0], [101,'Short', 1],[102,'Medium', 2],[103,'Long', 0],[104,'Short', 1],[105,'Medium', 2]]
intervention_df = pd.DataFrame(data, columns = ['ID', 'Intervention', 'GradeLevel'])

lookup_data = [[100, 0], [101, 1],[102, 2],[103, 0],[104, 1],[105, 2], [106, 0], [107, 0],[108, 2],[109, 1]]
lookup_df = pd.DataFrame(lookup_data, columns = ['ID', 'GradeLevel'])


df= pd.merge(intervention_df.assign(y='Yes'), lookup_df, on=['ID', 'GradeLevel'], how='outer')
df.loc[df.y.isnull(), 'y'] = 'No'


    ID Intervention  GradeLevel    y
0  100         Long           0  Yes
1  101        Short           1  Yes
2  102       Medium           2  Yes
3  103         Long           0  Yes
4  104        Short           1  Yes
5  105       Medium           2  Yes
6  106          NaN           0   No
7  107          NaN           0   No
8  108          NaN           2   No
9  109          NaN           1   No

Upvotes: 2

Related Questions