Graph adjusted line?

[Anna Stubbendorff]

Do you know if it is possible to graph a line, adjusted for confounders? In this example it looks like there is a rise in haemoglobin with dietary greenhouse gas emissions, but this is because men usually have higher haemoglobin and higher greenhouse gas emissions than women. In this example – can I draw a line adjusted for sex? Does the first line in my code actually do anything?

code:
regress HGB Total_CO2 i.SEX
twoway (lfit HGB Total_CO2) || (lfit HGB Total_CO2 if SEX==0) || (lfit HGB Total_CO2 if SEX==1), ytitle(HGB) xtitle("Dietary GHGE (kg/day)")

[Andrew Musau]

help margins

regress HGB Total_CO2 i.SEX
margins SEX, at(Total_CO2=(1/25))
marginsplot

[Anna Stubbendorff]

Thank you! Now i get 2 lines, one for men, one for women. Can I get only one line?

[Andrew Musau]

Omit the gender variable if you want the overall margin.

[Anna Stubbendorff]

Thank you for your expertise so far. Hopefully this is my last question this time.

Why does my line look exactly the same with this codes, when my regression coefficient is not the same? It seems like it is not graphing the coefficient with covariates in the model.


regress HGB AGE Total_CO2
twoway lfit HGB Total_CO2

regress HGB Total_CO2
twoway lfit HGB Total_CO2

[Andrew Musau]

As illustrated in #2, margins will give you the adjusted predictions and marginsplot will plot them. twoway is a graphing command and does not take into account any commands ran before it. The graphs are the same because the two calls to twoway are the same.

twoway lfit HGB Total_CO2

[Anna Stubbendorff]

So it is not possible to plot a simple regression line based on your coefficient with more variables in the model? Do I have to do it manually?

[Nick Cox]

Check out added variable plots.

help avplots 

search favplots

Not quite what you're asking for, but often as or more useful.

[Anna Stubbendorff]

I have checked the help section but I cannot find the information I need. See explanation below:

[Nick Cox]

I would just be guessing if I tried again with a code suggestion. But my guess is that you want a graph for each predictor that will be less informative than an added variable plot, so I will stop there.

[Anna Stubbendorff]

Thank you for your time! I am new to Stata and I am sorry if I have misunderstood something. I will keep learning :-)