Interpreting Stata Output

[Taiba Chau]

Hi, I just need some help interpreting the output of my results:

This is evaluating the impact of the policy imposed in Germany on the well-being of ethnic and non-ethnic individuals.
Therefore the coefficient 0.0306374 estimates the difference in the mental wellbeing of ethnic individuals compared to non-ethnic individuals.

I'm trying to understand whether there result is significant or not?

[Maxence Morlet]

The coefficient UK#Wave#Ethnic (or Germany#Wave#Ethnic, which is it?) is very far from statistically signifcant (p value way above 0.1). The probability vale (column P>|t|) telles you the lowest possible significance level at which you can reject the null hypothesis, here that the coefficient is not statistically different from 0.

Would you mind showing us your equation?

[Jared Greathouse]

Part of the reason we ask to not post screenshots and instead just posting the actual table is because posting the full table allows us to get a clear picture of what the result was. I know what the tables look like and how they're structured, but for the sake of others who may not be on their computers (like me right now), it makes it clearer for everyone to just post the table itself and not the screenshot.


Anyways, rant over. To answer the question, no, the coefficient we see here is not statistically significant. I'm not sure why you'd hang your proverbial hat on this given its various shortcomings, but no, your coefficient is not significant by the conventional interpretation.


EDIT: Your 3 way interaction with the UK and Germany are the exact same. As are the SE and the stat, even the confidence intervals. I think your problems may go a little bit deeper than the size of your p-values.

[Taiba Chau]

Sorry the UK screenshot was not correct. But the I was wondering what exactly do you look at to see if it is statistically significant or not? Is it the p-value or confidence interval?

[Maxence Morlet]

You can look at both. In #2 I worte in brackets "p-value way above 0.1" so in this case I looked at the p-value.

If 0 is comprised in the 95% confidence interval, it will not be significant at 5%.

[Marcos Almeida]

Just a short comment: probably, the interaction terms are not improving the model, since they provided large p-values. I would stick to a (parsimonious) model without the interaction terms and check the GOF stats.

[Carlo Lazzaro]

Taiba:
as an aside to previous excellent advice, please note that the FAQ ask to post what you typed, too. From your post #1 I can only guess that you went -regress-, but I'm not clear why you interacted country with (data) waves.
Are you dealing with a panel dataset? A survey? A repeated cross-sectional study?
Have you considered how difficult is to retrieve meaningful results from three-way interactions and how tricky is to disseminate them?
In addition, have you performed the necessary tests to investigate whether the functional form of your regressand is correct?