Pulling variable names when using pandas and statsmodels

Question

I'm trying to access the names of variables from the results generated by statsmodels. I'll elaborate more after the example code.

import scikits.statsmodels.api as sm
import pandas as pd
data = sm.datasets.longley.load()
df = pd.DataFrame(data.exog, columns=data.exog_name)
y = data.endog
df['intercept'] = 1.
olsresult = sm.OLS(y, df).fit()

This summary output includes the variable names. When you call something like olsresult.params it returns the following:

In [21]: olsresult.params
Out[21]: 
GNPDEFL           15.061872
GNP               -0.035819
UNEMP             -2.020230
ARMED             -1.033227
POP               -0.051104
YEAR            1829.151465
intercept   -3482258.634596

Now what I'm curious about doing is creating something like a dictionary with the variable name as a key and the parameter value as the value. So, something like {'GNPDELF':15.0618, 'GNP':-0.035819} and so on. If it's impossible to do this, is there any other way to access the variable name and value individually?

Answer 1

It's always worth trying the obvious.. :^)

In [14]: olsresult.params
Out[14]: 
GNPDEFL           15.061872
GNP               -0.035819
UNEMP             -2.020230
ARMED             -1.033227
POP               -0.051104
YEAR            1829.151465
intercept   -3482258.634597

In [15]: dict(olsresult.params)
Out[15]: 
{'ARMED': -1.0332268671737328,
 'GNP': -0.035819179292614578,
 'GNPDEFL': 15.061872271452557,
 'POP': -0.051104105653539733,
 'UNEMP': -2.0202298038172479,
 'YEAR': 1829.151464613984,
 'intercept': -3482258.6345966831}

See also the .to_dict() method of Series objects.