CODEWITHSUNDEEP
pythonpandasrecursionrows
TylerNG
TylerNG

Reputation: 941

python performing recursion on rows in df

I want to perform the following calculations with my df. I've done this in Excel but I'm not sure how to do this using Python. Here's the calculation 

      initial|  recursion begins
Name  | 1    | 2        | 3        | 4   ...
------------------------------------------------
A     | A1   | A2       | A3       | A4         <--given 
a     | a1=0 | a2=c1*A2 | a3=c2*A3 | a4=c3*A4   <--calculation (1st cell always = 0)
c     | c1=1 | c2=c1-a2 | c3=c2-a3 | c4=c3-a4   <--calculation (1st cell always 1)

Here's my example:

df

Name    1    2     3    4     5
----------------------------------
A       0   .125  .286  .25  .333
B       0   0     0    .5    -

the output would be:

Name    1    2     3     4    5
----------------------------------
A       0   .125  .286  .25  .333
Ax      0   .125  .25   .156 .156  
Ay      1   .875  .625  .469 .313
B       0   .1    0     .25    -
Bx      0   .1    0     .225   -
By      1   .9    .9    .675   -

Thank you!

Upvotes: 1

Views: 92

Answers (1)

BENY
BENY

Reputation: 323226

Let me explain a little bit .

a2=c1*A2;c2=c1-a2, then c2=c1-a2=c1-c1*A2=c1(1-A2),

c3 should be c3=c2(1-A3)=c1(1-A2)(1-A3), that is where(1-df).cumprod(1), come from 

df=df.set_index('Name') 
df1=(1-df).cumprod(1)
df2=df1.shift(1,axis=1).mul(df)

pd.concat([df,df1,df2],keys=['','x','y']).fillna({'1':0})


Out[769]: 
          1      2        3         4         5
  Name                                         
  A     0.0  0.125  0.28600  0.250000  0.333000
  B     0.0  0.000  0.00000  0.500000       NaN
x A     1.0  0.875  0.62475  0.468562  0.312531
  B     1.0  1.000  1.00000  0.500000       NaN
y A     0.0  0.125  0.25025  0.156187  0.156031
  B     0.0  0.000  0.00000  0.500000       NaN

Upvotes: 2

Related Questions