Re: t-value

[Jonathan Stickel]

Jaroslav Hajek wrote:
> On Mon, Apr 28, 2008 at 3:26 PM, Jonathan Stickel <address@hidden> wrote:
> > Jonathan Stickel wrote:
> >
> > > David Bateman wrote:
> > >
> > > > Jonathan Stickel wrote:
> > > >
> > > > > OK, easy enough now that I know how.  The changeset file is attached.
> > > > >
> > > > > Maybe a "how to submit changes" paragraph could be added to the "Help
> > > > > Wanted" part of the webpage?
> > > > >
> > > > > Regards,
> > > > > Jonathan
> > > > I always thought CDF stood for "cumulative distribution function"..
> > > >
> > > > D.
> > > You are right!  Simple word choice errors as I was typing.  Here is
> > another changeset, or you can just make the trivial changes yourself if you
> > prefer.
> > > Jonathan
> >  It doesn't look like my changeset(s) with improved help text for tcdf and
> > tinv has been checked in.  Is there anything else that I need to do?
>
> Probably not, just to be patient  - sometimes the patches just pile
> up. Fixes usually have priority over enhancements (for good reasons).
> But anyway, if you discover a bug in a patch you've posted and you see
> it wasn't yet applied (John always posts a notification), I guess it's
> better to redo the original changeset and post again, so that there is
> only one changeset to apply. You can do this using Mercurial by doing
> "hg rollback" to undo the commit, fix the bug, and commit & export
> again.

OK, here is a comprehensive patch that avoids the typo of the previous 
patches that I sent to the list.

I have a couple comments about octave development.  Although I am a 
small-fry contributor to open-source projects, mostly octave related, I 
think I generally understand the process.  To maintain control, not 
everyone is allowed to check code in directly.  That is fine.  Instead, 
there is usually a process where everyone can contribute by submitting 
bugs, etc.  For octave this is done by submitting patches (now 
changesets) to the lists.  JWE and others do a great job responding 
appropriately and checking stuff in.  However, this is not as 
transparent as it could be, and occasionally some submissions get missed 
or forgotten, and the contributor doesn't always know the status of 
his/her contribution (take this thread as an example).  I prefer a web 
interface bug submission process where all can see and keep track of the 
submissions.  I realize effort would be needed to set up this type of 
system, but I think it is worthwhile.  What do others think?

Jonathan

# HG changeset patch
# User address@hidden
# Date 1209420081 21600
# Node ID 56e58e0596ec8d1dc1c09bfcd7aa107c2ec1caa9
# Parent  6848970153ba5e2e1b59284bd258d91abd2e0e96
updated help for tcdf and tinv

diff -r 6848970153ba -r 56e58e0596ec scripts/statistics/distributions/tcdf.m
--- a/scripts/statistics/distributions/tcdf.m   Fri Apr 25 12:17:44 2008 -0400
+++ b/scripts/statistics/distributions/tcdf.m   Mon Apr 28 16:01:21 2008 -0600
@@ -18,9 +18,9 @@
 
 ## -*- texinfo -*-
 ## @deftypefn {Function File} {} tcdf (@var{x}, @var{n})
-## For each element of @var{x}, compute the CDF at @var{x} of the
-## t (Student) distribution with @var{n} degrees of freedom, i.e.,
-## PROB (t(@var{n}) <= @var{x}).
+## For each element of @var{x}, compute the cumulative distribution
+## function (CDF) at @var{x} of the t (Student) distribution with
+## @var{n} degrees of freedom, i.e., PROB (t(@var{n}) <= @var{x}).
 ## @end deftypefn
 
 ## Author: KH <address@hidden>
diff -r 6848970153ba -r 56e58e0596ec scripts/statistics/distributions/tinv.m
--- a/scripts/statistics/distributions/tinv.m   Fri Apr 25 12:17:44 2008 -0400
+++ b/scripts/statistics/distributions/tinv.m   Mon Apr 28 16:01:21 2008 -0600
@@ -18,9 +18,11 @@
 
 ## -*- texinfo -*-
 ## @deftypefn {Function File} {} tinv (@var{x}, @var{n})
-## For each component of @var{x}, compute the quantile (the inverse of
-## the CDF) at @var{x} of the t (Student) distribution with parameter
-## @var{n}.
+## For each probability value @var{x}, compute the the inverse of the
+## cumulative distribution function (CDF) of the t (Student)
+## distribution with degrees of freedom @var{n}.  This function is
+## analagous to looking in a table for the t-value of a single-tailed
+## distribution.
 ## @end deftypefn
 
 ## For very large n, the "correct" formula does not really work well,