Confidence and prediction interval for dependent variable after MLR

[Ashtone lazarre]

Hello ,
I'm running a multiple regression analysis , and i'm looking to find confidence and prediction intervals for the dependent variable given fixed values of independent variables
Is there a commande to do this using stata 13?
Thank you

[Clyde Schechter]

I'm not sure I understand your request, but it sounds like -predict- will do this. Pay particular attention to the xb and stdp options. Then you can get the confidence intervals by adding and subtracting 1.96*stdp to the xb result. (I'm assuming you are working with a large sample; if your sample is small, then replace the 1.96 by the appropriate figure for the t-distribution with the right number of degrees of freedom.)

[Bruce Weaver]

Hi Ashtone. Here is a thread from last year you may find helpful. Using the stdf option for predict was the key to getting the individual prediction interval Keowmani Thamron was looking for.

• https://www.statalist.org/forums/for...argins-command

Cheers,
Bruce

[Marcos Almeida]

Maybe you could take a look at -margins - command.

For example, getting the predicted mean value in a given model for, say, 8, 12 and 16 (you may consider as years of education):

margins, at(year_educ=(8 12 16))

[Bruce Weaver]

Ashtone lazarre is asking about both confidence intervals and prediction intervals. I believe she is making the same distinction that is shown on slides 54 & 55 here:

• https://www.slideshare.net/vermaumes...ssion-38653351

Both margins and predict with the stdp option give confidence intervals (as defined on slide 54); but in order to get prediction intervals (slide 55), one must use predict with the stdf (standard error of forecast) option. Richard Williams demonstrated this in post #11 here:

• https://www.statalist.org/forums/for...argins-command

HTH.

[Rich Goldstein]

I see that in post #2, Clyde suggests using the stdp option while in posts #3 and #5, Bruce suggests using the stdf option; these are different and the OP should read the Stata documentation carefully to see which is appropriate in her case

[Clyde Schechter]

Actually, a close reading of the original post suggests that she is interested in both of these errors, as recommended by Bruce.