How to perform breusch-pagan test in statsmodels?

Question

I want to use breusch_pagan test in statsmodels, but i facea strange error:

print(het_breuschpagan(resid=lr.resid,exog_het=df['iq']))
  File "C:\Users\aleks\PycharmProjects\statistics\venv\lib\site-packages\statsmodels\stats\diagnostic.py", line 810, in het_breuschpagan
    nobs, nvars = x.shape
ValueError: not enough values to unpack (expected 2, got 1)

linear model:

lr = sm.OLS.from_formula('lw80~age80+iq+school80+expr80',df).fit()

Dataset:

  rns rns80  mrt mrt80 smsa  ...  expr80  tenure  tenure80     lw   lw80
0  no    no   no   yes  yes  ...  10.635       0         2  5.900  6.645
1  no    no   no   yes  yes  ...  11.367       2        16  5.438  6.694
2  no    no   no   yes  yes  ...  11.035       1         9  5.710  6.715
3  no    no   no   yes  yes  ...  13.089       1         7  5.481  6.477
4  no    no  yes   yes  yes  ...  14.402       3         5  5.927  6.332

I would like to test wether there is a heteroskadicity assuming that the variable leading to heteroskadicity is iq and only iq, but I don't know what parameters to use so that there are no errors

Answer 1

The bresuschpagan implementation in statsmodel compares the residuals to the X dataset used in your ols regression, so that would be

het_breuschpagan(resid=lr.resid, exog_het=df[['age80','iq','school80','expr80']])