[rkward-devel] RKWard-devel Digest, Vol 94, Issue 9

[Thomas Friedrichsmeier]

Hi,

On Saturday 20 September 2014 16:43:48 Aaron Batty wrote:
> K, that installed and ran okay, but the weirdest thing was that the
> regression dialog (only, it seems) has weird stretched texted. It's a
> graphical error, but I can't figure out why it wouldn't do the same thing
> everywhere. When I try to screencap the window, the problem goes away. None
> of the other analyses do it! So weird.

weird, indeed. No idea. Is the whole dialog affected, or only a particular 
region of it?
 
> By the way, the thing that started this whole discussion, including an
> option to just save the lm object to the workspace (a la the ANOVA plugin)
> is not actually addressed by the SVN version of the regression plugin. It
> only saves the fitted values, when what is really needed is the residuals.

Well, true. I was thinking, we can't possibly offer individual "save-boxes" 
for all the data in the lm-object, and so I kept it really frugal, 
deliberately. After all, going from fitted values to residuals is not so 
hard... (When I did my first steps in R, the hardest part in that area was 
figuring out how to make sure input and fitted values remain paired up in case 
of NAs - but the plugin does handle that).

But on second thought, all other info in the lm-object is either stuff we 
already print to the output, or a copy of data that was fed into the model as 
input. And residuals are important enough to provide them really conveniently. 
Now in SVN, and we'll probably prepare more bundles for testing, soonishly.

Now I'll just have to work on making the <saveobject> element somewhat 
smarter, but that's probably something for after 0.6.2.

Regards
Thomas