How to generate a normal probability plot of residuals after linear regression?

[Tom Hsiung]

For some important reasons, after doing a linear regression analysis, a residual plot and a normal probability plot of residuals must be done to check if the data meets the prerequisites of linear regression (see following). The residual plot could be generate by syntax of rvpplot. How about normal probability plot of residuals?

[Nick Cox]

I suspect you're violating copyright in posting that and StataCorp would rather you didn't do that.

You need a better textbook in that assuming a normal distribution is less crucial than that extract implies.

rvpplot is not what is usually called a residual plot without qualification; that's more likely to be rvfplot

Otherwise consider

. sysuse auto, clear
(1978 Automobile Data)

. regress weight price

      Source |       SS           df       MS      Number of obs   =        74
-------------+----------------------------------   F(1, 72)        =     29.42
       Model |  12791822.9         1  12791822.9   Prob > F        =    0.0000
    Residual |  31302355.5        72  434754.937   R-squared       =    0.2901
-------------+----------------------------------   Adj R-squared   =    0.2802
       Total |  44094178.4        73  604029.841   Root MSE        =    659.36

------------------------------------------------------------------------------
      weight |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       price |   .1419244   .0261645     5.42   0.000     .0897663    .1940824
       _cons |   2144.459   178.5954    12.01   0.000     1788.436    2500.483
------------------------------------------------------------------------------

. predict res, res

. qnorm res

[Marcos Almeida]

If I understood your question correctly, you may use - predict - to create the residuals, then apply - qnorm - as well as - pnorm plots.

With regards to the debate on the "incontournable" assumptions for linear regression, you may find several thought-provoking threads in this forum.

P.S: Crossed with Nick's reply.