[RkWard-devel] [Updated] Distribution plots.

[Prasenjit Kapat]

Hi,

I have made additional changes which Thomas has added to svn 
(under_development.pluginmap). The chagelog is as follows (** == fresh 
changes). There are a few "polls"/opinions at the end of this mail. Kindly 
comment. Of course, comments/suggestions other than the polls are also 
appreciated.

A. Menu Change:
--------------------------
[1] Wilcoxon Rank Sum distribution is discrete, so moved it from the 
continuous to discrete collection.

B. External Changes:
--------------------------------
[1] Added a checkbox each for "log"/"log.p" and "lower.tail" parameters to the 
density/cdf plots for all the listed distribution.

[2] Added the "preview" tag to all the distribution plots. 

[3] Added a "Parameter" frame around the parameter values.

[4] Changed the Minimum/Maximum labels to Lower/Upper quantiles. After all, R 
calls "x,q" as vector of quantiles.

[5] Changed the heading of the Outputs. For example from "Plot density dnorm" 
to "Normal density function". For the discrete ones, used "mass function" 
instead.

**[6] "Choose type of paramerization" radio now enables / disables the 
corresponding spinboxes for the Negative Binomial Distribution.

**[7] "Choose the type of function plot" radio now enables/disables the "Lower 
Tail" checkbox.

**[8] For Tukey distribution, there is no density plot. So, the I have 
disabled the "Choose the type of function plot" radio and added a text 
underneath.

C. Internal Changes
------------------------------
[1] Converted all the internal plotting function to the style suggested by 
Thomas: 
              function(x) <? getRK ("function");?> (x,...)

[2] Added "Non-centrality" parameter wherever it was applicable.

[3] Assigned proper data type (integer/real) for the various parameters and 
variables used.

[4] Provided bounds (lower and/or upper, as the case may be) to the parameters 
and variables of the "spinboxes".

**[5] Changed all the php files so that getRK is called exactly once for each 
variable. echo ($VAR) is used in place of getRK ("VAR") for any subsequent 
calls.

**[6] "log = TRUE" and "log.p = TRUE" are used instead of "log = 1" to specify 
the log scale for plotting.

**[7] Similarly "lower.tail = FALSE" is used instead of "lower.tail = 0."

Polls / Opinions:
-----------------------
[1] For discrete distributions with bounded support (like binomial, wilcoxon) 
or the case where the support is depended (bounded one way or the other) on 
the parameter specification (like hypergeometric, negative binomial), I was 
planning to remove the "Lower/Upper quantile specification" and calculate 
them based on the parameter specification (as is actually done for the 
wilcoxon). But I was not completely sure of this approach. I was hoping some 
suggestions from the community.

[2] For discrete distributions which plot type is better? type = "h" or type 
= "p".

[3] Should "Beta distribution" or just "Beta" be used in the menu entries? (I 
am just repeating this from my last email)