CODEWITHSUNDEEP
rstatisticslinear-regression
R_D
R_D

Reputation: 11

Understanding the output of my Ramsey RESET test

I am new to R, and doing a replication study where I need to check if their regression holds for the classical assumption for OLS. For the specification assumption, I am doing the Ramsey RESET test, here is my code:

simple_model <- lm(deploy ~ loggdppc + natoyears + milspend + lagdeploy + logland + logcoast + lag3terror + logmindist)

resettest(simple_model, power=2, type="regressor", data = natopanel)

Here is my output:

RESET = 2.0719, df1 = 6, df2 = 355, p-value = 0.05586

Since the P-value is (albeit slightly) above 0.05, does this mean that it 'passes' the RAMSEY test? Or is there an issue of missing variables? I still have not gotten quite the hang of these interpretations. This model does not include all their variables, as they are testing for a specific hypothesis.

Thank you for your help!

Upvotes: 1

Views: 9535

Answers (2)

Daken Engmann
Daken Engmann

Reputation: 1

I would review your response that using 'fitted' gives you a p-value of 0.4047.

Remember that the purpose of this test is to see if in your model we should use quadratic or higher power terms. The null hypothesis is that quadratic (or higher power) is better. If we reject the null in this case then we are saying that statistically first order terms are better.

When you get a p-value of less than 0.05 this means that you reject the null hypothesis - this means that statistically your model is better with first order terms. When you got 0.05568 you should've investigated further if you needed more than first order terms in your model.

When you added in 'fitted' etc the p value was even higher. This means that there is even less certainty that first order terms are sufficient.

It is REALLY important to know what you are doing with the values "fitted", "regressor" and "principal".

Upvotes: 0

Otto K&#228;ssi
Otto K&#228;ssi

Reputation: 3083

According to Wikipedia:

"[The intuition of Ramsey RESET] test is that if non-linear combinations of the explanatory variables have any power in explaining the response variable, the model is misspecified in the sense that the data generating process might be better approximated by a polynomial or another non-linear functional form"

It tests whether including higher degree polynomials of your explanatory variables -- in your example 2nd degree due to power=2 -- have any additional explanatory power. In essence, you test whether the 2nd-degree terms of your regressors are jointly significantly different from zero.

Suppose you use 5% as your cut-off for significance. In that case, you (barely) fail to reject the null hypothesis that including the 2nd-degree terms improves the fit over a linear model.

Upvotes: 1

Related Questions