Error when trying to extract the P-value from a rolling OLS regression

[Charalampos Karagiannakis]

Hi. I have been using command 'rolling' to estimate a rolling OLS regression of Y on X1. I have been trying to store the rolling constant estimates together with the underlying p-values. Specifically, I am running the following command:

rolling alfa=_b["_cons"] pval=r(table)["pvalue" ,"_cons"], window(20): regress Y X1

I am getting the following error message.

type mismatch
error in expression: r(table)
r(109);

Could anyone bring some light? Thank you.

[William Lisowski]

What release of Stata are you using?

[Jared Greathouse]

I'm just guessing here, but it looks like your trying to store the entire table matrix into the pval stat...... maybe the solution would be to do the equivalent of

Pvalue=_p["var"]

by the way, I've no experience with this specific prefix, I was just guessing

[Charalampos Karagiannakis]

Thank you both.
William Lisowski, I am working on StataMP 16.1.
Jared Greathouse, the allowed underscore variables are only _b and _se. I tried using _p in any case, but didn't work.

[William Lisowski]

This is a little subtle. If you look at the output of help rolling and click on exp_list in the command syntax, you will see the output of help exp_list which requires careful reading to figure out that if you are not using _b or _se but rather a standard Stata expression like r(table)["pvalue" ,"_cons"] you will need to enclose that expression in parentheses. So your command should work if you add parentheses as in

rolling alfa=_b["_cons"] pval=(r(table)["pvalue" ,"_cons"]), window(20): regress Y X1

My apologies for sending you down a dead end by asking about your version; I was away from my computer and was unable to confirm that the problem is not version-related until now.

[William Lisowski]

it looks like your trying to store the entire table matrix into the pval stat

No, that is standard syntax for subscripting matrices using row and column names.

. sysuse auto, clear
(1978 automobile data)

. regress price weight

      Source |       SS           df       MS      Number of obs   =        74
-------------+----------------------------------   F(1, 72)        =     29.42
       Model |   184233937         1   184233937   Prob > F        =    0.0000
    Residual |   450831459        72  6261548.04   R-squared       =    0.2901
-------------+----------------------------------   Adj R-squared   =    0.2802
       Total |   635065396        73  8699525.97   Root MSE        =    2502.3

------------------------------------------------------------------------------
       price | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
      weight |   2.044063   .3768341     5.42   0.000     1.292857    2.795268
       _cons |  -6.707353    1174.43    -0.01   0.995     -2347.89    2334.475
------------------------------------------------------------------------------

. matrix list r(table)

r(table)[9,2]
            weight       _cons
     b   2.0440626  -6.7073534
    se   .37683413   1174.4296
     t   5.4243032  -.00571116
pvalue   7.416e-07   .99545897
    ll   1.2928575    -2347.89
    ul   2.7952677   2334.4753
    df          72          72
  crit   1.9934636   1.9934636
 eform           0           0

. // extract the pvalue for the constant
. display %9.3f r(table)["pvalue","_cons"]
    0.995

. // in earlier versions this syntax was needed
. display %9.3f r(table)[rownumb(r(table),"pvalue"),colnumb(r(table),"_cons")]
    0.995

.

[Charalampos Karagiannakis]

This is a little subtle. If you look at the output of help rolling and click on exp_list in the command syntax, you will see the output of help exp_list which requires careful reading to figure out that if you are not using _b or _se but rather a standard Stata expression like r(table)["pvalue" ,"_cons"] you will need to enclose that expression in parentheses. So your command should work if you add parentheses as in

rolling alfa=_b["_cons"] pval=(r(table)["pvalue" ,"_cons"]), window(20): regress Y X1

My apologies for sending you down a dead end by asking about your version; I was away from my computer and was unable to confirm that the problem is not version-related until now.

Thank you very very much. This example should be added to the 'rolling' help file!

[Jared Greathouse]

Oh yes William Lisowski , you're right it is. I didn't recognize it as such at first since I usually subscript by number.