[rkward-devel] [rkward-users] Proposition of changes on crosstabulation plugin

[Thomas Friedrichsmeier]

Hi,

I'll reply to your points in a different order of writing. 

> PD2: I sent this message to the users list not to the developers list,
> because some one could test the plugin with this changes and make
> suggestions, and I have see that many people want to get contingency tables
> with more information, well there is a method :)

I'll take this to rkward-devel, instead. While of course this is correct, the 
main point of having separate lists for users and developers is to avoid 
flooding all users' inboxes with the details of the development (which often 
generates a lot of noise). We do regularly invite subscribers of rkward-users 
for testing, before new releases are made.

*This* mail still has rkward-users in CC, so nobody will be left wondering 
about the reply, but please send further follow-ups to rkward-devel.

> I think that we could make a repository of user contribuited plugins, or
> forks to the existing plugins.

We have the beginnings of that since RKWard 0.5.5. See Settings->Configure 
RKWard->Plugins->Install or uninstall add-on plugin packs. 
http://rkward.sourceforge.net/documents/devel/plugins/external_plugins.html 
for documentation on how to create such "plugin packs". Currently, this 
repository is hosted on Meik's server, so you will need to get in touch with 
Meik for details on uploading.

On Thursday 31 March 2011, Andrés Necochea wrote:
> Some weeks ago I sent a message to the developement list with a proposition
> of changes on the crosstabulation plugin of rkward. In a conversation winth
> Thomas, I propose a function that use ftable to add statistic to an R
> contingency table. Thomas make some fixes to this function.
> 
> I attached the file containing the plugin. At now, doesn't have many
> features, only uses the bind.tables function to include the statistics in
> the main table. It will include % of row, % of column, % of total, an chisq
> expected.

I only had a cursory look so far. I have some changes in mind. Will try to 
give you detailed feedback within a few days. As a first detail, here's another 
slight simplification I made to the bind.tables() function, shortly after our 
discussion about it. I believe it does all that we want, but test for 
yourself:

  bind.tables <- function (...) {
     tables <- list (...)
     output <- unlist (tables)
     dim (output) <- c (dim (tables[[1]]), length (tables))
     dimnames (output) <- c (dimnames (tables[[1]]), list (statistic=names 
(tables)))
     ftable (output, col.vars=2)
  }

> I'm working in the bind.tables function to work on a matrix with more than
> two dimensions.

Perhaps the above version of bind.tables() can be extended to more dimensions 
more easily, too.

> I tryng to make some changes on the dialog of the plugin to include a tab
> with the table options.

Alternatively, consider adding those options to the first tab (which already 
has the chisquare option). But of course this is just a detail, and can be 
changed, easily, at any time.

> Beside, I'm planing to change the rk.results function to include a method
> to print a ftable class object.

Out of curiosity: What will be the main difference(s) with respect to
  rk.print(some_ftable_object)
?

> If some one have a suggestion, or a correction to this plug in, please tell
> me.

Again, I'll try to give you more detailed feedback within a few days.

Regards
Thomas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/rkward-devel/attachments/20110331/600ea113/attachment.sig>