Margins plots and significance

[William Salazar]

Hello Stata Folk,

I'm running a panel regression with squared interaction terms. I want to confirm my interpretation of my marginsplot - I hope/think I am on the right track!

Data:
Firm level data N approx 2000, T=14
y= financial performance
x=ESG performance
sector: categorical, 10 sectors.

Regression:

xtreg y x c.x#c.x sector#c.x sector#c.x#c.x, fe vce(cluster company)

Margins:

margins sector_name if inlist(sector,30), dydx(x) at(x=(20(10)90))

n.b. I have many sectors, just plotting one.

Resulting graph:



Interpretation:
- The only significant effects are at x = 20, 30, and 40.
- I interpret this as there is a negative relationship between ESG(x) and financial performance (y) for low levels of ESG BUT this negative relationship reduces in magnitude as ESG increases.
- At x=>50 there is no relationship.

Question 1:
- Is this interpretation sound?

Question 2:
- I have a lot of plots for a lot of sectors. Could I just plot the values that were significant? On the one hand this would make it easier to look at lots of data on a big panel of charts more easily. On the other, I appreciate I am dealing with a continuous variable in X. See below for an example of the type of plot I am thinking for each sector:


Any insight gratefully received.

[William Salazar]

I don't know why I cannot see my post in the Stata general forum. Strange. I reply to my own post in the hope it makes it appear!