This is possibly too specific of a question, but here goes:
I am running a regression of log of abortions on selected dependent
variables which include a series of 2-way & 3-way interactions. Afterwards,
I test linear combinations using lincom. When including the last 2 3-way
interactions & specifying the robust option only, lincom spits out the
following error: MATRIX NOT POSITIVE DEFINATE. I can see no reason why this
should occur. I can manually compute the lincom estimate that should result
using SE's & the variance/covariance matrix. This problem does not occur if
I use both robust & cluster as options or if I use neither robust nor
cluster. Does anyone have ideas?
LINCOM FAILS:
xi3: reg abrtLn i.parity*waiver i.parity*tanf ueRate_cur ueRate_lag1 maxAfdc
i.state*ttrend1 i.state*ttrend2 popnLn*ttrend1 popnLn*ttrend2
i.parity*popnLn i.state*popnLn*ttrend1 i.state*popnLn*ttrend2
[aw=abrtTot] if cell==1, robust
LINCOM WORKS:
xi3: reg abrtLn i.parity*waiver i.parity*tanf ueRate_cur ueRate_lag1 maxAfdc
i.state*ttrend1 i.state*ttrend2 popnLn*ttrend1 popnLn*ttrend2
i.parity*popnLn
[aw=abrtTot] if cell==1, robust;
--
Danielle H. Ferry
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
