This regards an old post that posed the question:
Tom Hodgess wrote:
"The problem is the (apparent?) inability to produce moving range
process behavior (a.k.a. "control") charts with individuals data in
the
package "qcc" (v. 2.0). I have also struggled with the same limitation
in package "IQCC" (v. 1.0).
The package "qAnalyst" (v. 0.6.0) provides an option to produce a
moving range chart with individuals data. "
[It does, but the look of the qAnalyst chart is very different from the
qcc chart, and the qcc charts generally provide more information.]
To which Filthy McNasty replied [on 16 Jan 2010]:
-----
The
moving range chart has the same control limit factors (viz., D3, D4) as
the range chart, so one should be able to feed an
appropriately-constructed matrix into qcc(type = "R"). In other words,
if one can transform the original individual observations 1..m into 1..
(m - 1) rational subgroups of size two so that the range of the
rational subgroup i equals the moving range of the individual
observations i and i+1, it'll fool qcc's range chart into plotting the
moving ranges (MR) as the range (R), computing the center line as the
average MR instead of the average R, and computing the control limits
as D3*MR and D4*MR instead of D3*R and D4*R.
My solution for qcc, using Montgomery's viscosity data in
"viscosity.txt":
batch viscosity phase
1 34.05 setup
2 34.40 setup
3 33.59 setup
...
observations <- read.table("viscosity.txt", TRUE)
require(qcc)
attach(observations)
# identify "observation" and "batch" columns as providing
the
# observations and unit numbers
observation <- qcc.groups(viscosity, batch)
# number of batches in phase I ("setup")
m <- length(phase[phase == "setup"])
# extract the batch numbers to display for the phase II ("monitoring")
data
setupBatches <- as.character(batch[phase == "setup"])
monitoringBatches
<- as.character(batch[phase == "monitoring"])
# plot xbar chart
obj <- qcc(data = observation[1,1:m],
type = "xbar.one",
newdata = observation[1,-(1:m)],
labels = setupBatches,
newlabels = monitoringBatches,
axes.las = 3,
chart.all = TRUE,
title = "Individuals chart for aircraft primer paint
viscosity", xlab = "Batch",
ylab = "Value (viscosity units)")
setupBatches <- setupBatches[-1]
numberOfBatches <- length(observation)
# intermediate matrix to coax qcc into plotting a moving range chart
#
# observations x[i], x[i + 1] appear in adjacentObservations[i]
# whose range equals the moving range |x[[i]] - x[[i + 1]]|
#
adjacentObservations <- matrix(nrow = length(observation) - 1, ncol 2)
adjacentObservations[,1] <- observation[1,1:numberOfBatches - 1];
adjacentObservations[,2] <- observation[1,2:numberOfBatches];
# plot MR chart
obj <- qcc(data = adjacentObservations[1:m - 1,],
type = "R",
newdata = adjacentObservations[-(1:m - 1),],
labels = setupBatches,
newlabels = monitoringBatches,
axes.las = 3,
chart.all = TRUE,
title = "MR chart for aircraft primer paint viscosity",
xlab = "Batch",
ylab = "Moving range (viscosity units)")
-----
For the life of me, I can't get the above to work. I use "qcc" in
a
production environment, and Xmr charts would be valuable to have, and
they should look like the charts produced by qcc. Here is my work
around [note, I am not a programmer. I'm sure it is ugly, and there
are no doubt more elegant ways to do this.]
temp.data looks like this [lot number and value. Lot number header
left off for other manipulations, lot numbers needed for some
manipulations and not others. Adjacent data (see below) does not need
lot numbers]:
pH
08-10538 7.6
08-11736 7.5
08-12453 8.1
08-12557 7.9
08-13202 7.4
# read table, with header
z<-read.table('temp.data',1)
# create 'group' column, populate with sequence of integers
z2<-cbind(z, group=seq(1,length(z[,1]),1))
# use length(z[,1]) -1 units of data to create adjacent data
z3<-rbind(z2, c(z2[-1,]))
# renumber the adjacent data's groups
z3[(length(z[,1]) + 1):(length(z3[,1])),2] <- z3[(length(z[,1]) + 1):
(length(z3[,1])),2] - 1
# remove the last member of the initial data, so the groups all have
# two values
z4<-z3[-(length(z[,1])),]
# now group with qcc
obj<-qcc.groups(z4$pH, z4$group)
# now plot using the R graph function of qcc
xmr<-qcc(obj, type="R", title="Xmr Chart")
I've checked it against my manual calculations, and it seems to work
the way I want it. If you're trying to produce Xmr charts with
"qcc",
this is one example of how to do it.
--
jim smith <n0oct at sbcglobal.net>
