Python Pandas average based on condition into new column

Question

I have a pandas dataframe containing the following data:

matchID    server    court    speed
1          1         A         100
1          2         D         20         


        

Answer 1

With groupby, we can still use loc to select the intended parts that we want to replace but put the whole computation within a for loop from df.groupby("matchID").

for id, subg in df.groupby("matchID"):       
    df.loc[df.matchID==id, "meanSpeedCourtA13"] = (subg
              .where(subg.server.isin([1,3])).where(subg.court == "A").speed.mean())
    df.loc[df.matchID==id, "meanSpeedCourtD13"] = (subg
              .where(subg.server.isin([1,3])).where(subg.court == "D").speed.mean())

Specail thanks to @Dark to point it out that I was hard coding groupby.

For loc, it can be used to select values based on information from 2 axes: rows and columns. By convention on the documentation, the sequence to put information is rows first and columns second. For example, in df.loc[df.matchID==id, "meanSpeedCourtD13"], df.matchID==id is about selecting rows that have matchID being id and that "meanSpeedCourtD13" specifies a column we want to look into.

Side notes about calculating mean:

• for each group subg
• where(subg.server.isin([1,3])) then filter out server not in [1 ,3].
• where(subg.court == "A") further to do filtering on court.
• finally call mean to compute mean from speed.

---

As an alternative, you can use np.where to assign values to each matchID in [1, 2]. This works only for binary matchID. It is roughly the same speed with the groupby method above tested on my computer. To save space, we only demonstrate with "meanSpeedCourtA13" column.

# First we calculate the means
# Calculate mean for Group with mathcID being 1
meanSpeedCourtA13_ID1 = (df[df.matchID==1].
                 where(df.server.isin([1,3])).where(df.court == "A").speed.mean())    
# Calculate mean for Group with matchID being 2
meanSpeedCourtA13_ID2 = (df[df.matchID==2].
                 where(df.server.isin([1,3])).where(df.court == "A").speed.mean())
# Use np.where to allocate values to each matchID in [1, 2]
df["meanSpeedCourtA13"] = np.where(df.matchID == 1,
                                   meanSpeedCourtA13_ID1, meanSpeedCourtA13_ID2)

For np.where(condition, x, y), it will return x if condition is met, y otherwise. See np.where for documentation.