<br><br><div><span class="gmail_quote">2007/1/12, Thomas Friedrichsmeier <<a href="mailto:thomas.friedrichsmeier@ruhr-uni-bochum.de">thomas.friedrichsmeier@ruhr-uni-bochum.de</a>>:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
On Friday 12 January 2007 09:54, I. Soumpasis wrote:<br>> 2) Indentation: The next line after an opening brace ('{') should always be<br>><br>> > indented by one Tab (until the closing '}'). This makes reading the code
<br>> > structure much easier. Also, maybe you could replace the spaces that are<br>> > used<br>> > to indent the panel.cor()-block right now with tabs, as well.<br>><br>> You 're right. Done.<br>
<br>Sorry, I wasn't precise: The indentation I was worried about is the<br>indentation of the generated code, not the indentation as visible in the .php<br>file. I.e. inside the preprocess () function, you'd unindent everything by
<br>two tabs, inside the printout (), and cleanup () functions by one tab. Then,<br>the generated code should look "clean". (See the new section on that in the<br>plugin documentation:<br><a href="http://rkward.sourceforge.net/documents/devel/plugins/phptemplate.html#policyformatting">
http://rkward.sourceforge.net/documents/devel/plugins/phptemplate.html#policyformatting</a>)</blockquote><div><br><br>My bad. I took off one tab from the first 2 tabs.<br></div><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
> 3) It would probably be useful to have the "method", and "use" options to<br>><br>> > cor<br>> > and cor.test as well. You can copy the relevant sections from the<br>> > correlation
<br>> > matrix plugin. One way to make this work, would be to extend the<br>> > panel.cor function with those two paramenters, then modify the call to<br>> > pairs to:<br>> ><br>> > pairs(x,
lower.panel=panel.smooth, upper.panel=function (x, y)<br>> > panel.cor(x, y, use, method))<br>><br>> I have done it with a different method, don't know if it is ok. I applied<br>> use, method to the cor and
cor.test functions. Tested the method and works.<br>> Is this ok or should I try the method proposed?</blockquote><div><br>I did that at the function panel.cor like this <br>panel.cor <- function(x, y, digits=<? getRK ("digits"); ?>,
cex.cor, use="<? getRK ("use"); ?>", method="<? getRK ("method"); ?>") <br><br>and then main function is <br><br>pairs(x, lower.panel=panel.smooth, upper.panel=panel.cor
)<br><br></div><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">I think there is one advantage to the method I proposed: I would make the<br>
panel.cor () function truly reusable as is. All parameters would be set at<br>only one single point: The function call itself. The user would not have to<br>modify the function body itself to re-run e.g. with method "spearman" instead
<br>of "pearson". I think in this case this method would not introduce too much<br>additional complexity, and be useful at the same time, so I think it's<br>better.<br><br>> 4) The "digits" and "prefix" parameters to
panel.cor are effectively<br>> unused,<br>><br>> > currently. I don't really understand, what the prefix parameter is for,<br>> > but<br>> > maybe you could also add a GUI option for the digits parameter (that, or
<br>> > remove it to keep the code simple).<br>><br>> I gave an option to the user to select precision (1,2,or 3 digits - Sould<br>> we do something else here?). Removed prefix.<br><br>It's probably nicer to use a <spinbox type="integer" min="0" max="5"
<br>initial="2"> instead of a <radio> for this task. After all, it's a numerical<br>selection.</blockquote><div><br><br>Done.<br></div><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
> 5) You wrap up everything in a function rk.temp.cor.graph (). I'm not<br>><br>> > entirely<br>> > sure, whether I like or dislike this (so far we did not use this).<br>> > Anybody have an opinion already?
<br>><br>> There is not a function that can do this to my knowledge so I made this<br>> one using the example. This is the way I have implemented it in R and just<br>> pasted here. One good thing is that you have to delete only this one
<br>> function stored in R (and your data). One other that someone could pass<br>> scripts or functions already written in a very easy way. I don't know.<br>> Should I change that?<br><br>The alternative would be to rename "
panel.cor" to "rk.temp.panel.cor", then<br>change<br> rk.temp.cor.graph (rk.temp.x)<br>to:<br> pairs (x, lower.panel=panel.smooth, upper.panel=rk.temp.panel.cor)<br>(or considering 3) to):<br>
pairs (x, lower.panel=panel.smooth, upper.panel=function (x, y)<br>rk.temp.panel.cor (x, y, digits=<? getRK ("digits"); ?>, use="<? getRK<br>("use"); ?>", method="<? getRK ("method"); ?>"))
<br><br>However, since I'm not quite sure myself, just leave everything inside the<br>rk.temp.cor.graph () function for now, until I've made up my mind.<br><br>> 7) The idea of scaling the text the r is printed in according to the value
<br>><br>> > of<br>> > r is definitely interesting, but I think it should be possible to turn it<br>> > off.<br>><br>> Done.<br><br>Since this is a yes/no question, a <checkbox> would be better suited than a
<br><radio>. Considering 3), you might want to add scale as a parameter to<br>panel.cor, too. Also, for this type of selection, you might want to set the<br>option value to "TRUE" or "FALSE", instead of "yes" or "no". Then you can
<br>simplify a statement like:<br><br> if("<? getRK ("scale"); ?>"=="yes") [...]<br>to<br> if(<? getRK ("scale"); ?>) [...]</blockquote><div><br>Done and used TRUE/FALSE.
<br><br>I have to look at the documentation. This was a first try to see what can I do. <br></div><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
> 1. I added a note with what the asterisks mean.<br><br>Very useful, indeed.<br><br>> 2. I believe that the plugin should be placed under the same menu with<br>> correlation matrix, I mean something like this.<br>
><br>> Analysis -- Correlation -- Correlation matrix<br>><br>> -- Correlation graph<br><br>Not sure, yet. Anybody else have an opinion?<br><br>Regards<br>Thomas</blockquote><div>
<br>Added also info like this<br><br>Method: pearson<br>Exclusion: pairwise.complete.obs<br>Precision: 3 digits<br>Scale text: TRUE<br><br>on the graph. <br></div><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
-------------------------------------------------------------------------<br>Take Surveys. Earn Cash. Influence the Future of IT<br>Join SourceForge.net's Techsay panel and you'll get the chance to share your<br>opinions on IT & business topics through brief surveys - and earn cash
<br><a href="http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV">http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV</a><br><br>_______________________________________________
<br>RKWard-devel mailing list<br><a href="mailto:RKWard-devel@lists.sourceforge.net">RKWard-devel@lists.sourceforge.net</a><br><a href="https://lists.sourceforge.net/lists/listinfo/rkward-devel">https://lists.sourceforge.net/lists/listinfo/rkward-devel
</a><br><br><br><br></blockquote></div><br>