pandas - LOOP univariate rolling window regression on entire DF Python

Question

I have a dataframe of 24 variables (24 columns x 4580 rows) from 2008 to 2020.

My independant variable is the first one in the DF and the dependant variables are the 23 others.

I've done a test for one rolling window regression, it works well, here is my code :

import statsmodels.api as sm

from statsmodels.regression.rolling import RollingOLS

import seaborn

seaborn.set_style('darkgrid')

pd.plotting.register_matplotlib_converters()

 

x = sm.add_constant(df[['DIFFSWAP']])

y = df[['CADUSD']]

rols = RollingOLS(y,x, window=60)

rres = rols.fit()

params = rres.params

r_sq = rres.rsquared

Now, what i want to do, i'd like to do a loop to regress (rolling window) all the dependant variables of the DF (columns 2:24) on the independant variable (column 1) and store the coefficients and the rsquareds.

My ultimate goal is to extract Rsquareds and Coefficients and put them in dataframes(or lists or whatever) and then graph them.

I'm new to Python so I'd be very gratefull for any help.

Thank you!

question from:https://stackoverflow.com/questions/65836299/loop-univariate-rolling-window-regression-on-entire-df-python

Answer 1

Can you throw it all in a loop and store the results in some other object like a dict?

Potential solution:

data = {}
for column in list(df.columns)[2:]:  # iterate over columns 2 to 24

    x = sm.add_constant(df[column])
    y = df[['CADUSD']] ## This never changes from CADUSD, right?
    rols = RollingOLS(y, x, window=60)
    rres = rols.fit()
    params = rres.params
    r_sq = rres.rsquared
    # Store results from each column's fit as a new dict entry
    data[column] = {'params':params, 'r_sq':r_sq}

results_df = pd.DataFrame(data).T