Singularity problems with RandomEffectsEstimator

[danielbr2]

Hello,

I'm trying to replicate an example from Wooldridges Introductory Econometrics textbook, but I'm having problems with the RandomEffectsEstimator. In the example below, the variable year seems to cause the error LinearAlgebra.SingularException(6). If left out, the problem disappears. Is there a bug in the code, or am I missing something?

using WooldridgeDatasets, GLM, DataFrames, Econometrics

wagepan = DataFrame(wooldridge("wagepan"))

# RandomEffectsEstimator gives singularity problems:
reg_re = fit(RandomEffectsEstimator,
    @formula(lwage ~ educ + black + hisp + exper + (exper^2) +
                     married + union + year),
    panel=:nr, time=:year,
    wagepan,
    contrasts=Dict(:year => DummyCoding()))

Without error I would expect something like Pythons linearmodels :

import wooldridge as woo
import pandas as pd
import linearmodels as plm

wagepan = woo.dataWoo('wagepan')
reg_re = plm.RandomEffects.from_formula(
    formula='lwage ~ educ + black + hisp + exper + I(exper**2) +'
            'married + union + C(year)', data=wagepan)
results_re = reg_re.fit()
table_re = pd.DataFrame({'b': round(results_re.params, 4),
                         'se': round(results_re.std_errors, 4),
                         't': round(results_re.tstats, 4),
                         'pval': round(results_re.pvalues, 4)})
print(f'table_re: \n{table_re}\n')

# table_re: 
#                     b      se       t    pval
# C(year)[1980]  0.0234  0.1514  0.1546  0.8771
# C(year)[1981]  0.0638  0.1601  0.3988  0.6901
# C(year)[1982]  0.0543  0.1690  0.3211  0.7481
# C(year)[1983]  0.0436  0.1780  0.2450  0.8065
# C(year)[1984]  0.0664  0.1871  0.3551  0.7225
# C(year)[1985]  0.0811  0.1961  0.4136  0.6792
# C(year)[1986]  0.1152  0.2052  0.5617  0.5744
# C(year)[1987]  0.1583  0.2143  0.7386  0.4602
# educ           0.0919  0.0107  8.5744  0.0000
# black         -0.1394  0.0480 -2.9054  0.0037
# hisp           0.0217  0.0428  0.5078  0.6116
# exper          0.1058  0.0154  6.8706  0.0000
# I(exper ** 2) -0.0047  0.0007 -6.8623  0.0000
# married        0.0638  0.0168  3.8035  0.0001
# union          0.1059  0.0179  5.9289  0.0000

[Nosferican]

Hi, I suspect it may have to do something with the perfect multicollinearity. Thanks for reporting the reproducible example. I will try to take a look at it this week.

[danielbr2]

Did you have a chance to look into it? Appreciate your feedback!

[Nosferican]

I am so sorry. I haven't. Got assigned a release schedule and seminar in the last few weeks, which came out this week and presented the seminar this morning. I will try to catch my breath and look at it this weekend. Thanks for following up I haven't forgotten but been quite busy in this last window.