[rkward-devel] Jarque-Bera test plugin

Germán Márquez Mejía manchito at gmail.com
Tue Oct 2 02:22:52 UTC 2007


Hi:

As I'm not sure if this message was correctly sent yesterday, I send it again.

By the way, nice bug and nice solution that of the 64 bits problem.

Thomas Friedrichsmeier escribió:
> Hi,
>
> On Saturday 29 September 2007, Germán Márquez Mejía wrote:
>> As I've been kind of free this week I decided to take the challenge and
>> try with a simple plugin for a frequently used test (at least in my case).
>
> great!
>
>> Clearly this is NOT FOR 0.4.8. Of course, any suggestion, modification
>> or even rejection is welcome. I'm not an expert and there are several
>> things I ignore about making good plugins and so, but the initiative is
>> here.
>
> A very nice first try. A couple of points below (don't let it scare you, all
> of this is minor stuff):
> 1) We have a policy of adding new plugins to under_development.pluginmap,
> first, to make sure we don't accidentally release them without decent
> testing. Also, we add them to this page:
> http://rkward.sourceforge.net/wiki/index.php?title=Development_Status_of_Plugins
> to keep track of the development status.

Yes, I missed that, sorry.

> 2) The plugin indeed has only one option so far, so on a certain level it
> makes sense not to use a tabbook with two tabs for that. However, it's also a
> good idea to design similar plugins to look similar. The other distribution
> tests use two tabs, so I suggest, this one should as well. Also the "Show
> Length" option should probably be "Show Length and NAs" for the same reason
> (see below).

Done

> 3) The issue about NAs not handled by the test is one we could easily solve
> for the user. Just offer an additional option "Exclude NAs".

Done, but I'm not quite sure about offering such an option because, if
data does have NAs and the option is unchecked, the test fails. Instead,
I'd put a note of advise saying that all missing values will be ignored,
unless that option could be useful for the user to prevent running the
test on such data.

> This option would do something like (after counting the length and NAs):
>       var <- var[!is.na(var)]

You're right, it's very easy. Done.

> 4) The "parameters" argument for rk.header() is meant for user selected
> parameters, not results (arguably, the degrees of freedom is sort of a
> parameter, but you get the point). So, the df should rather be included in
> the results, not in the header. The "Exclude NAs" option would be a good
> candidate for inclusion in the header.

Done. I did it because I didn't like the idea of a column of 2s, but it
may be necessary if you need to copy i.e. only one row of the results.

> I've taken care of 1), and added your work to the SVN repository (by the way,
> I also added your translation, thanks!) for after 0.4.8. Perhaps you would
> like to take on the others yourself to get some more practice.

Thanks. I'm glad not everything was wrong :-) I'll try not to come up
with all this few days before a release in the future (hope there's such
a future - I mean, for me). It's just that tomorrow I return to my
everyday tasks and wanted to leave things in a decent status until 0.4.8
passes.

I attach the .php, .xml and .rkh updated files.

Regards,

--
Germán
-------------- next part --------------
A non-text attachment was scrubbed...
Name: jb_test_002.tar.gz
Type: application/x-gzip
Size: 1641 bytes
Desc: not available
URL: <http://mail.kde.org/pipermail/rkward-devel/attachments/20071001/8bd2f8c0/attachment.gz>


More information about the Rkward-devel mailing list