Re: [R] bwplot puts the bars in the wrong place

Sat, 17 Apr 2010 09:10:05 -0700 


On Apr 17, 2010, at 11:01 AM, James Rome wrote:


Ah,

Well, I thought I put up just the data needed. Sorry, was wrong file.
And yes I am still climbing up the very steep learning curve hill. I
attach a better data set. If it gets stripped, it is at
http://dl.dropbox.com/u/537118/gdf.txt
1. I put the gdf$tt~gdf$OnHour because in the rest of my code, there are 
other (previous and bigger) incarnations of this data frame with the
same names. Or does the call automatically assume that the variable
names are from the frame in the data= statement? How am I to know?
Yes, R does "assume" such. Formula methods are a rather fundamental concept in R. Your readings should have exposed you to that concept. 


2. The plot you enclosed is still not correct. The 0 time things are
plotted at 1. And I believe that there are no data at 3 am. That is why 
I asked the question, and I think this is also the key to the problem.
Had that been clearly stated earlier, I might have realized my solutions were not complete. 


3. I did create factors as Jun suggested, and they are in OnHFact. But
using OnHFact instead of OnHour also gives the wrong plot.
I'm assuming that Ehlers' posting of a few minutes ago was the correct solution, i.e, that you needed a more complete specification of levels of the factor? 



The only thing that put the correct data on the correct hours was to
call xyplot instead of bwplot, with panel.bwplot in the panel function. 

Sorry for being so dense,  but I really find it much harder to read R
documentation, than say, Java documentation. And I have 10 R books
including R Graphics, which did not shed light on this issue either.

Thanks for the help,
Jim Rome

On 4/17/10 9:52 AM, David Winsemius wrote:

After unzipping that file, http://dl.dropbox.com/u/537118/gdf.zip , I
got a structure object that took what seemed to be an inordinately
long time to eval-parse wehn assigned to gdf,  but it did eventually
produce:


str(gdf)

'data.frame':    2656 obs. of  21 variables:
$ OnDate        :Class 'Date'  num [1:2656] 14222 14221 14334 14424
14519 ...
$ FltOrigDt     : chr  "12/9/2008" "12/8/2008" "3/31/2009"
"6/29/2009" ...
$ MkdCrrCd      : chr  "DL" "DL" "DL" "DL" ...
$ MkdFltNbr     : int  742 1517 1517 1517 1517 1699 1065 1777 1777
1777 ...
$ DprtTrpnStnCd : chr  "DEN" "JAX" "JAX" "JAX" ...
$ ArrTrpnStnCd  : chr  "ATL" "ATL" "ATL" "ATL" ...
$ ActualOutLocal: POSIXct, format: "2008-12-09 01:04:00" "2008-12-08
05:37:00" "2009-03-31 05:32:00" ...
$ ActualOffLocal: POSIXct, format: "2008-12-09 01:47:00" "2008-12-08
06:11:00" "2009-03-31 05:43:00" ...
$ ActualOnLocal : POSIXct, format: "2008-12-09 06:10:00" "2008-12-08
06:56:00" "2009-03-31 06:33:00" ...
$ ActualInLocal : POSIXct, format: "2008-12-09 06:18:00" "2008-12-08
07:02:00" "2009-03-31 06:39:00" ...
$ ArrivalGate   : Factor w/ 100 levels "A01","A02","A03",..: 2 2 2 2
2 2 2 2 2 2 ...
$ DepartureGate : chr  "C44" "A7" "A7" "A10" ...
$ Flight        : chr  "DAL742" "DAL1517" "DAL1517" "DAL1517" ...
$ OnHour        : chr  "6" "6" "6" "6" ...
$ OnDateTime    : POSIXct, format: "2008-12-09 06:10:00" "2008-12-08
06:56:00" "2009-03-31 06:33:00" ...
$ Runway        : Factor w/ 10 levels "08L","08R","09L",..: 2 4 4 7 4
7 8 8 1 7 ...
$ Delay         : int  15 8 NA NA NA NA NA NA NA NA ...
$ TaxiTime      :Class 'difftime'  atomic [1:2656] 480 360 360 240
360 420 300 300 540 780 ...
 .. ..- attr(*, "units")= chr "secs"
$ passurdt      : num  0 0 0 0 1 0 0 0 0 -7 ...
$ OnHFact       : Ord.factor w/ 24 levels "0"<"1"<"2"<"3"<..: 7 7 7 7
7 8 8 13 13 12 ...
$ tt            : num  8 6 6 4 6 7 5 5 9 13 ...



On Apr 16, 2010, at 9:37 PM, James Rome wrote:


On 4/16/2010 8:27 PM, Jun Shen wrote:Jim,

Try this,

bwplot(tt~as.factor(OnHour),data=gdf,......)

Jun Shen from Millipore

I already tried using a factor, and the data set I enclosed had
gdf$OnHFact which was already a factor. It gave the same wrong plot.

What did work was to call xyplot instead of bwplot, and to use
panel.bwplot in the panel function:

      hrs = seq(0, 23, 1)
     hrlabs = as.character(seq(0,23,1))
      g = xyplot(gdf$tt~gdf$OnHour |gdf$Runway, data=gdf,


This is a puzzling way to invoke xyplot and may have unforeseen
dangers. Generally when one uses a data argument to a plot function,
one does not also include that dataframe name in the formula terms on
either the LHS or the RHS.


ylab="Taxi
time (min)", main=title, xlab="Hour of day",
          xlim=c(-1, 24), scales=list(x = list(rot=90, cex=.6,
alternating=c(3,3,3,3),
          at=hrs, labels=hrlabs
          )),
          panel=function(x, ...) {
              panel.grid(h = -1, v = 24)
              panel.bwplot(x, horizontal=FALSE, col="black",...)

          }
      )
      print(g)
I took your initial code for a bwplot call and took out the extraneous 
dataframe name in the formula terms and put in your axis code (with
more sensible formatting) and got what appears to be your desired plot: 

bwplot( tt~ OnHour |Runway, data=gdf,
        scales=list(x = list(rot=90, cex=.6,
                              alternating=c(3,3,3,3),
                              at=hrs, labels=hrlabs
                    )        ),
        horizontal=FALSE)

Plot attached.
But I do not understand why this makes a difference. It has something to 
do with the fact that there are no data for some of the hours.


I have not been able to figure out what you meant by "bars in the
wrong place". You could have been more forthcoming about what you saw
as the problem for the Readers of the list. My guess is that it has
something to do with not following the specified conventions for
arguments to the bwplot formula method.
The difference between the calls is either a bug, or it should be in 
the documentation somewhere obvious. I spent a week on this.


To report a bug you need a lot more specifics about your system and
better efforts at isolation to minimal cases.



Thanks,
Jim Rome

Thanks,
Jim Rome


<gdf.txt>


David Winsemius, MD
West Hartford, CT

______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

Reply via email to