Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# st: mi impute: VCE is not positive definite

 From Lena Sperling To statalist@hsphsun2.harvard.edu Subject st: mi impute: VCE is not positive definite Date Sun, 1 Jul 2012 13:49:34 +0200

```Dear all,

I'm trying to impute some household data sets. I have a problem with
the wage variable, which is continuous between 2.2 and 7707.91.

Variable Obs Mean Std. Dev. Min Max
mwage 3908 318.6525 295.8956 2.20226 7707.91

But when I impute it either using pmm or regress I get the message:
mi impute: VCE is not positive definite
The posterior distribution from which mi impute drew the
imputations for pci is not proper when the VCE estimated from the
observed data is not positive definite.  This may happen, for
example, when the number of parameters exceeds the number of
observations. Choose an alternate imputation model.
error occurred during imputation of contract healthins ocusec industry
firmsize_l firmsize_u pci edulevel2 mwage on m = 1

My code is: mi impute chained (pmm) contract healthins ocusec industry
firmsize_l firmsize_u pci edulevel2 (regress) mwage= union idh urb
reg01 reg02 hhsize head gender age ed_mod_age literacy lb_mod_age
njobs unitwage if lstatus==1 & empstat==1 [pweight=wgt], add(10)
augment force noisily

And the mi xeq 0: code results in:

m=0 data:
-> regress mwage

Source |       SS       df       MS              Number of obs =    3935
-------------+------------------------------                     F(
0,  3934) =    0.00
Model |           0     0           .                 Prob > F
=       .
Residual |   343172406  3934  87232.4368           R-squared     =  0.0000
Total |   343172406  3934  87232.4368           Root MSE      =  295.35

------------------------------------------------------------------------------
mwage |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
_cons |   318.3377   4.708327    67.61   0.000     309.1067    327.5686
------------------------------------------------------------------------------

Any ideas? There are only 10.8 pct missing data, so it should not be a problem.