[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: -modeldiag- available on SSC
Thanks to Kit Baum, a -modeldiag- package has been
added to SSC, superseding the following packages:
-modeldiag- contains these programs,
each with a corresponding .hlp:
The same names occur as above, with the
exception of -racplot-, which is new on SSC.
In addition, there is an overarching help,
-modeldiag- is a bundle of graphics programs
that may be run after fitting a regression-type
command. Programs are written for Stata 8, except
that in most cases a previous version written for Stata
7 is also included in the package. There is nothing
for Stata 6 or earlier versions.
Numbering conventions are that "2" added to
the name indicates a program offered
as a generalisation of the corresponding
official Stata command and that "7" added
to the name indicates a program written for
There is a strong bias here towards plots
useful with responses that are continuous
or can be treated as such. Lest this
be labelled "continuous chauvinism" and
met with categorical contempt let me underline
that Scott Long and Jeremy Freese maintain
various substantial -spost*- packages, including
not only graphics but various other commands.
Official Stata has programs with similar intent
documented at [R] regression diagnostics.
In no case does any of those run except after
-regress-, or possibly -anova-. Such restrictions are not
built into the programs in -modeldiag-,
except that -anovaplot- does require a prior -anova-.
Official Stata also has commands like -twoway lfit-,
-twoway qfit-, -twoway fpfit-, which although intended
as exploratory tools, may also be run after a model
It is difficult to know of all related work, but
Phil Ender and Michael Mitchell from UCLA have
also written several similar programs.
This package originated under Stata 7 and programs
written at that time are frozen as is, with names like
-anovaplot7-. I rewrote most of the programs for Stata 8
in early 2003, and have now again revised them all.
In most cases, the revision is minor or at least should
be upward compatible with the version previously on SSC,
but in a few cases I have had second thoughts about the syntax.
-indexplot- is a case in point. This may
break a few .do files, or habits, but I hope does not
cause real inconvenience. Maintaining the
Stata 8 versions kept on SSC until just now would just
be a complication too far.
Similarly, I hope that abolishing the
individual packages in favour of a single
bigger one causes no inconvenience and indeed
is convenient for anyone interestd.
Archives need maintenance, as Kit's almost daily
work makes clear!
Kit Baum, Denis de Crombrugghe, Phil Ender, Ken Higbee,
and Andy Sloggett also kindly commented on earlier
versions of programs now in -modeldiag-.
* For searches and help try: