[R] Colour filling in panel.bwplot from lattice

[Deepayan Sarkar]

On Wed, Nov 3, 2010 at 4:25 PM, Rainer Hurling <rhurlin at gwdg.de> wrote:
> Am 03.11.2010 10:23 (UTC+1) schrieb Deepayan Sarkar:
>>
>> On Wed, Nov 3, 2010 at 4:11 AM, Dennis Murphy<djmuser at gmail.com>  wrote:
>>>
>>> Hi:
>>>
>>> I don't know why, but it seems that in
>>>
>>> bwplot(voice.part ~ height, data = singer,
>>> main = "NOT THE RIGHT ORDER OF COLOURS\n'yellow' 'blue' 'green' 'red'
>>> 'pink' 'violet' 'brown' 'gold'",
>>> fill=c("yellow","blue","green","red","pink","violet","brown","gold"))
>>>
>>> the assignment of colors is offset by 3:
>>>
>>> Levels: Bass 2 Bass 1 Tenor 2 Tenor 1 Alto 2 Alto 1 Soprano 2 Soprano 1
>>> fillcol<- c("yellow","blue","green","red","pink","violet","brown","gold")
>>>
>>> In the above plot,
>>>
>>> yellow ->  Bass 2  (1)
>>> blue ->  Tenor 1     (4)
>>> green ->  Soprano 2  (7)
>>> red ->  Bass 1 (10 mod 8 = 2)
>>> pink ->  Alto 2 (13 mod 8 = 5)
>>> etc.
>>>
>>> It's certainly curious.
>>
>> Curious indeed. It turns out that because of the way this was
>> implemented, every 11th color was used, so you end up with the order
>>
>>> sel.cols<-
>>> c("yellow","blue","green","red","pink","violet","brown","gold")
>>> rep(sel.cols, 100) [ seq(1, by = 11, length.out = 8) ]
>>
>> [1] "yellow" "red"    "brown"  "blue"   "pink"   "gold"   "green"
>>  "violet"
>>
>> It's easy to fix this so that we get the expected order, and I will do
>> so for the next release.
>
> Thank you for this proposal. We are looking forward for the next release :-)
>
> We frequently have to colour selected boxes to be able to compare special
> cases over different panels.
>
>> Having said that, it should be noted that any vectorization behaviour
>> in lattice panel functions is a consequence of implementation and not
>> guaranteed by design (although certainly useful in many situations).
>> In particular, it is risky to depend on vectorization in multipanel
>> plots, because the vectorization starts afresh in each panel for
>> whatever data subset happens to be in that panel, and there may be no
>> relation between the colors and the original data.
>
> Thank you for the warning.
>
>> One alternative is to use panel.superpose with panel.groups=panel.bwplot:
>>
>> bwplot(voice.part ~ height, data = singer, groups = voice.part, panel
>> = panel.superpose, panel.groups = panel.bwplot, fill = sel.cols)
>
> This indeed works nice 'as a workaround'.

Actually, I would reiterate that this is the "right solution" and the
it's other fix that qualifies as a quick workaround (especially if you
are considering comparing things across multiple panels).

-Deepayan