在 pandas 数据帧上拆分和执行 function 的最有效方法

Question

I have been given a data frame that contains two measurements of a value (A and B) in rows and each column represents the measurements for sample.

Example below:

ID S1 S2 S3
M1_A 1 2 3 
M1_B 3 2 1
M2_A 1 2 3 
M2_B 3 2 1 

I need to calculate the ratio of B to A+B [ie (B/(A+B))] for each measurement each sample.

Result data frame example:

ID S1 S2 S3
M1 0.75 0.5 .25 
M2 0.75 0.5 .25

Currently I am reading in the file two lines at a time, checking that the ID's match (excluding the _A or _B), transforming the "lines" to vectors and then performing the calculation across to vector. On larger sample sets this gets extremely slow.

What is the most efficient way to do this using a library such as pandas?

All help appreciated!

Answer 1

This sounds like a classic groupby-aggregate problem. Pandas can handle the underscore in the ID column easily as well.

df['ID'] = df['ID'].str.split('_').str[0]
df = df.groupby('ID').agg(lambda x: x.values[-1]/x.sum())
print(df)

      S1   S2    S3
ID                 
M1  0.75  0.5  0.25
M2  0.75  0.5  0.25