[rkward-users] Pareto Analysis using RKward

[Thomas Friedrichsmeier]

On Friday 03 June 2011, Vikas Garud wrote:
> I could get nice Pareto Chart, for the variable Defect code.  Is it
> possible to plot a Pareto Chart for, say, Total cost?  The attempt of
> plotting the same through RKward fails, in the sense that the number
> of times a particular cost value appears is used for plotting the
> Pareto chart.  I'd like a Pareto plot by defect codes based on Total
> cost.  Is this possible?  How?

Well, not via the GUI, currently. You'd have to transform the data like this:

    cumulated.costs <- by (data$total.cost, data$defect, FUN=sum)

Then plot this without further tabulation.

Looks like something that we could reasonably add to the GUI. 

> The data in this example to demonstrate Pareto Chart contains 8 defect
> codes and it is possible to make sense of the plot.  Some real data
> contains hundreds of records.  (There is file, somewhere on the
> internet - sorry, I could not find the link - CFHPareto.xls containing
> 11099 records in 37 defect codes.)  Doing a Pareto Analysis did give
> me nice coloured Pareto Chart, however, the legends are all mixed up
> and can not be read.  The utility of the Pareto chart is practically
> nil.  For such data, I would like to club categories beyond, say, 80%
> in cumulative total, to be clubbed in "Others".    Is it possible?
> How?

Well, if the problem is reading the legends, then you'd probably want to cut 
by number of categories, rather than a limit by cumulative total, no?

   cutoff <- 10
   if (length (cumulated.costs) > cutoff) {
       transformed <- sort (cumulated.costs, decreasing=TRUE)
       transformed <- c (transformed[1:cutoff], "Others"=sum 
(transformed[(cutoff+1):length(transformed)]))
   }

Might be a useul addition to the GUI, too.

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-users/attachments/20110603/81178c90/attachment.sig>