使用python statsmodels修复summary_col中的标签外生变量

Question

I want to produce regression tables like those yielded by summary_col (standard journal tables) but with custom explanatory variable labels.

Is there a way to change the row names saved in the model params attribute?

As of now I rename variables the closest I can to what I intend, but there ought to be a better way to do this.

Answer 1

Suppose you have done

reg = smf.ols(formula = "y~x1+x2+x3").fit()

I suggest (1) to have a dictionary where you hold all the relabeling: dic = {original_vname: new_name} a (2) a pair of useful functions:

def rename_vars(vname): 
    to_ret = vname
    for orig_vname in list(dic.keys()):
        if vname == 'original_vname':
            to_ret = dic['original_vname'] 
    return to_ret

and

def rename_ols(reg): 
    for i in range(len(reg)): 
        reg[i] = rename_vars(reg[i])

Then, just do:

rename_ols(reg.model.exog_names) 

And that's it. Once you call summary_col, the variables will show up with the new labels.