Line graph fitted values from categorical variables

[Solveig Delabroye]

Hi

I'm trying to create a line graph with predicted (fitted) values from a regression where the dependent variable is continuous (log earnings) but the explanatory variable is categorical (educational attainments).

The graph I get looks like a "nest", with every point connected to the others, and is unreadable.

I do not understand why this is happening.
While playing around with my data, I managed to have it appear "clean" once, but ever since this has been happening, even though I don't think I changed anything.

I use stata 12.

Here is my code:

reg logearn i.educrec, r
predict discretefitted
twoway (line discretefitted educyears)

Thank you so much!

[Nick Cox]

Your model fit doesn't lead me to suppose that the predictions are necessarily even monotonic in educyears -- which isn't a predictor.

But minimally you could insist on sorting

twoway (line discretefitted educyears), sort 

but as the predictions traverse a large region I think the problem is different.

Otherwise put, you need to explain the relationships between educrec and educyears for the question to be clear.

[Solveig Delabroye]

Thanks for your answer! "educyears" is just the continuous version of my education variable, so that the X-axis is labeled in years instead of arbitrary education levels.

The sort option is what I needed!