[Carlo Lazzaro]

Brian:
no, it is not (as you wrote, more conservative threshold of 30 is also reported; see page 2454 Stata .pdf manual).
In addition, if you have both linear and square terms for a given predictor, the -vif- value is, as expected, high (and unavoidably so):

. use "https://www.stata-press.com/data/r18/auto.dta"
(1978 automobile data)

. regress price i.foreign c.trunk##c.trunk

      Source |       SS           df       MS      Number of obs   =        74
-------------+----------------------------------   F(3, 70)        =      3.47
       Model |    82140034         3  27380011.3   Prob > F        =    0.0207
    Residual |   552925362        70  7898933.74   R-squared       =    0.1293
-------------+----------------------------------   Adj R-squared   =    0.0920
       Total |   635065396        73  8699525.97   Root MSE        =    2810.5

---------------------------------------------------------------------------------
          price | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
----------------+----------------------------------------------------------------
        foreign |
       Foreign  |    1177.11   768.6987     1.53   0.130    -356.0113    2710.231
          trunk |   355.4324   464.8687     0.76   0.447     -571.719    1282.584
                |
c.trunk#c.trunk |  -3.370961   16.64387    -0.20   0.840    -36.56612     29.8242
                |
          _cons |   1624.499   3099.483     0.52   0.602    -4557.225    7806.223
---------------------------------------------------------------------------------

. estat vif

    Variable |       VIF       1/VIF  
-------------+----------------------
   1.foreign |      1.16    0.864693
       trunk |     36.54    0.027367
     c.trunk#|
     c.trunk |     36.79    0.027184
-------------+----------------------
    Mean VIF |     24.83

.

[Brian Conway]

Carlo: one of the regression results for a variable was: -4.03e-07 (7.09e-07). Is it best to report this 'e' notation as is or is there more intuitive way of reporting the result? Would some transformation of the variable (e.g., log of the variable) make it more intuitive? The variable measures job growth (in thousands) over time. Thank you.

[Carlo Lazzaro]

Brian:
can't you scale it first to make its interpretation less problematic?

[Brian Conway]

Carlo: I was wondering how one can combine dealing with auto correlation (prais y x1 x2 x3, corc) with dealing with heteroskedasticity (regress y x1 x2 x3, robust). I added robust at the end of <prais y x1 x2 x3, corc> and this reported semirobust SEs rather than robust SEs. It also dropped one observation. Thank you.

[Brian Conway]

Carlo: I came across past literature suggesting that it is best to correct for auto correlation first and then heteroskedasticity in time series regression: Thus, in a regression (where Durbin-Watson statistic was .79 and the hettest was significant and chi2=6.34) I tried correcting for auto correlation using <prais x1 x2 x3, corc> and then ran <regress x1 x2 x3, vce(robust)> to correct for heteroskedasticity. I was wondering if this is correct? Thank you.

[Carlo Lazzaro]

Brian:
I do not think I can be more helpful, as I am not familiar with time series analysis.

[Brian Conway]

Thank you very much Carlo for your help with the previous posts.