Why is all variables not significant in panel treshold regression

[Hamid muili]

Dear all, I am using the xtendothresdpd package by Diallo (2020) to estimate the dynamic panel treshold regression with 37 countries and 20 years. All the independent variables aren't significant including the lag of dependent variable. All independent variables were chosen based on literature recommendations.
GDP=f( tax revenue, non tax revenue, debt, debt finace) could anyone suggest the appropriate way to use this package based on a paper or other important resources.

[Jared Greathouse]

All the independent variables aren't significant including the lag of dependent variable.

Why is that bad?

[Hamid muili]

Jared Greathouse , although I have re-estimated and I got a better result. However, when I tried to use the xtendothresdpdtest to test the significance of the treshold, it takes longer hours for the estimate to come up with a result. Please is there any solution to this?

[Jared Greathouse]

I don't know, you didn't provide any data or code so I won't comment on it, but I did ask you a philosophical question. You wrote

All the independent variables aren't significant including the lag of dependent variable. All independent variables were chosen based on literature recommendations.

and my question for you was

Why is that bad?

So I ask again, all the predictors were not significant. So what? You write that you get a better result after resetimation. What is your criteria for better? How can you tell it's better? I'm not trying to be mean, I should say, in asking this, I'm really asking you.

Beyond stats philosophy, in my experience, with ML estimators or other estimators that aren't based on least squares, sometimes, multiple hour estimation is the name of the game unfortunately. But before we even get to that, I do want to know why it's bad that all the predictors weren't significant. Assuming you coded everything right (which I can't judge here), maybe it's just true that the selected predictors aren't related to the outcome of interest?

Beyond this, what are you trying to model? Lots of context is missing for us to help you, but judging by the question, this is far afield from what I do, but I'm asking you these questions so others might have a better idea on what the goal is so they may be able to better help you.

[Joro Kolev]

All the variables that we study in our regression being insignificant, is a sign that we are not lucky :P.

The only thing that I see in your description which might be a problem is that your sample is very small. 37 countries are not that many.