Dear Gretl community,

I was estimating an OLS model using GUI and when I performed some tests on results I realized that heteroskedasticity tests - White's test, White's test (squared only), Breusch-Pagan, and Koenker - give a little bit confusing results (IMHO), look:

White's test for heteroskedasticity -
Null hypothesis: heteroskedasticity not present
Test statistic: LM = 51.4256
with p-value = P(Chi-Square(35) > 51.4256) = 0.0361713

White's test for heteroskedasticity -
Null hypothesis: heteroskedasticity not present
Test statistic: LM = 16.9311
with p-value = P(Chi-Square(14) > 16.9311) = 0.259869

Breusch-Pagan test for heteroskedasticity -
Null hypothesis: heteroskedasticity not present
Test statistic: LM = 7.01643
with p-value = P(Chi-Square(7) > 7.01643) = 0.427172

Breusch-Pagan test for heteroskedasticity -
Null hypothesis: heteroskedasticity not present
Test statistic: LM = 8.8388
with p-value = P(Chi-Square(7) > 8.8388) = 0.264438

With this results we aren't able to differentiate between the first and second results (and also the third and fourth results) because they are presented with the same title. I think it could be better if we have:

White's test for heteroskedasticity -

White's test for heteroskedasticity (squared only) -

Breusch-Pagan test for heteroskedasticity -

Breusch-Pagan test for heteroskedasticity (Koenker robust variant) -

What do you think?

Best,
Henrique C. de Andrade
Doutorando em Economia Aplicada
Universidade Federal do Rio Grande do Sul
www.ufrgs.br/ppge