将单独的函数应用于我的数据框，同时还对特定列求和

Question

I have a problem in which I have to summarize a lot of data and group it by three columns to summarize the data. A problem is that one specific column needs to have a specific formula applied to it as well.

My data looks like this:

Account Format  Network  Impressions Clicks Cost    Avg. position
Health1 Text     Search        2       0      0.5       1
Health1 Picture  Search        5       2      1        1.5
Health1 Picture  Search        1       2      3        2.4
Health1 Text     Search        1       0      0        2.3
Health1 Text     Display       2       0      0.7      1.7
Health2 Text     Display       0       0      0        3.3
Health2 Text     Display       2       2      4        3.3
Health2 Picture  Search        2       0      0        3.4
.....

So I need to group by the Account, Format and Network and summarize Impressions, Cost and Clicks like so, and summarize for each group:

Account Format  Network  Impressions Clicks Cost    Avg. position
Health1 Text     Search        3       0      0.5       x
Health1 Picture  Search        6       4      4         x
Health2 Text     Display       2       2      4         x
Health2 Picture  Search        2       0      0         x
.....

However, to calculate Avg. Position I need to apply a formula, my brain is kind of fried from working on stuff like this all day so any help would be a lifesaver. The Avg. Position column needs to have this formula applied to it:

sum(impressions*Avg. Position)/sum(impressions)

My attempt was to multiply the Avg. Position column by impressions on a row to row basis, summarize the column along with the others and then divide by the summed impressions. This is not correct apparently, as it returns values < 1, which is not a possible output in the context of the data I am using.

frame['Avg. position'] = frame.apply(lambda x: (x['Impressions']*x['Avg. position']), axis=1)
frame = frame.groupby(['Account', 'Format', 'Network'])['Impressions','Clicks','Cost','Avg. position'].sum().reset_index()

frame['Avg. position'] = frame.apply(lambda x: (x['Avg. position']/x['Impressions']) if x['Impressions'] > 0 else '', axis=1)

frame.to_csv(yesterday_date+'.csv', index=False)

Answer 1

The correct way to use apply to update your "Avg. position" column is this:

denominator = frame['Impressions'].sum()
frame['Avg. position'] = frame[['Impressions', 'Avg. position']].apply(lambda x: x[0] * x[1] / denominator, axis=1)

BUT ... with Series objects you can use element-by-element operations:

frame['Avg. position'] = frame['Impressions'] * frame['Avg. position'] / frame['Impressions'].sum()

which means you could also do something like this:

frame['Cost'] = frame['Cost'] / 1000000

I'm not exactly sure what you are trying to do with the groupby but it seems like you should have all the tools to figure it out now.