2008-08-16
maxControlsCap isn't reliable. Example: example(mahal.dist) ; minControlsCap(mhd2/(mhd2<2)).  The cap for '0' should be at least 1; but function returns 0.  -DONE
2008-03-21
To wishlist of methods for pcscore.dist, add objects from gbm package.  (Per suggestion from sguo@email.unc.edu)
2008-02 
Finish adding invertCov fct to makedist() ; see R/mahal.dist.Rnew
2008-01
Subsetting a vector of class optmatch causes loss of it treatment group attribute.  Write a subsetting method for that class to avoid this. DONE
2007-08-24
Following code crashed pscore.dist/makedist:
> pm2.low <- glm(update.formula(pscore.fmla, .~.*I(si.q1<=10)), [...]
> pd2.low <- pscore.dist(pm2.low, structure.fmla=~I(si.q1 <= 10))
(Cf anniAdmissions/stage1/match2007-08-24.Rnw)

Issue is that the model frame of pm2.low doesn't have v-ble 'si.q1',
rather 'I(si.qi <= 10)'.  Maybe pscore.dist should have an optional
data argument, just for evaluating structure.fmla?

2007-08-24 Chris Johnson tried to combine pscore.dist() with Harrell's
lrm() function, from Design package.  No go, b/c lrm() objects don't
have same value info as glm() objects.  Maybe pscore.dist should really
have methods for different kinds of fitted propensity models? (See also 2008-03-21 note)

2005
Consider adding refs to Harrell's matching functions find.matches and
matchCases.

Introduce omit.controls arguments to max.control.cap, min.control.cap.
Next step is to set it up to handle negative values of omit.controls
cap.  I seem to recall doing a bit of testing of omit.controls option
(in summer 2004).
