Hi, I have a subarray with 1567 spots replicated 4 times on each chip. I am trying to carry out the correlation analysis as in the Bob Mutant data example in the guide. I get as far as the cor$cor and then I get a correlation of 0.8 for every gene in the series. If I do the boxplot, I also get 0.8 for every tick on the axis. What am I doing wrong? It has defaulted to the initial estimate for correlation for every gene. I have tried altering "initial" in the dupcor.series function, to other values, and no matter what value I use, I get that value for every gene on the array. Perhaps loess for print tip is not effective for this number of genes per grid? thanks simon. RG <- read.maimages(files,source="genepix") Read CA05A16.gpr Read CA05B04.gpr Read CA05C04.gpr Read CA05C08.gpr Read CA05D08.gpr > RG <- backgroundCorrect(RG, method="none") > RG$genes <- readGAL() > RG$printer <- getLayout(RG$genes) > MA <- normalizeWithinArrays(RG) > targets <- readTargets("targets.txt") > boxplot(MA$M~col(MA$M),names=targets$Name) > MA <- normalizeWithinArrays(RG) > design <- designMatrix(targets, ref="mixed pool") Found unique target names: 4 day-1 mixed pool 4 day-2 4 day-3 4 day-4 4 day-5 > design 4 day-1 4 day-2 4 day-3 4 day-4 4 day-5 1 -1 0 0 0 0 2 0 -1 0 0 0 3 0 0 -1 0 0 4 0 0 0 -1 0 5 0 0 0 0 -1 > cor <- dupcor.series(MA$M,design,ndups=4) Loading required package: nlme Loading required package: lattice > cor $cor [1] 0.8 $cor.genes [1] 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 [19] 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 ...through to 1567 > cor$cor [1] 0.8 > boxplot(cor$cor.genes) -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: text/enriched Size: 1990 bytes Desc: not available Url : https://www.stat.math.ethz.ch/pipermail/bioconductor/attachments /20040103/a8dfd970/attachment.bin

You can't sensibly do any correlation analysis unless you have some degrees of freedom for error at the array level. You have 5 arrays and 5 coefficients in your linear model, i.e., only one array for each condition and no replication. You need at very least one more array than condition. Since the data provides no information about the correlation, you just get the initial value back. Gordon > Hi, > I have a subarray with 1567 spots replicated 4 times on each chip. I am > trying to carry out the correlation analysis as in the Bob Mutant data > example in the guide. I get as far as the cor$cor and then I get a > correlation of 0.8 for every gene in the series. If I do the boxplot, I > also get 0.8 for every tick on the axis. What am I doing wrong? It has > defaulted to the initial estimate for correlation for every gene. I > have tried altering "initial" in the dupcor.series function, to other > values, and no matter what value I use, I get that value for every gene > on the array. Perhaps loess for print tip is not effective for this > number of genes per grid? > > thanks > > simon. > > RG <- read.maimages(files,source="genepix") > Read CA05A16.gpr > Read CA05B04.gpr > Read CA05C04.gpr > Read CA05C08.gpr > Read CA05D08.gpr > > RG <- backgroundCorrect(RG, method="none") > > RG$genes <- readGAL() > > RG$printer <- getLayout(RG$genes) > > MA <- normalizeWithinArrays(RG) > > targets <- readTargets("targets.txt") > > boxplot(MA$M~col(MA$M),names=targets$Name) > > MA <- normalizeWithinArrays(RG) > > design <- designMatrix(targets, ref="mixed pool") > Found unique target names: > 4 day-1 mixed pool 4 day-2 4 day-3 4 day-4 4 day-5 > > design > 4 day-1 4 day-2 4 day-3 4 day-4 4 day-5 > 1 -1 0 0 0 0 > 2 0 -1 0 0 0 > 3 0 0 -1 0 0 > 4 0 0 0 -1 0 > 5 0 0 0 0 -1 > > cor <- dupcor.series(MA$M,design,ndups=4) > Loading required package: nlme > Loading required package: lattice > > cor > $cor > [1] 0.8 > > $cor.genes > [1] 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 > 0.8 0.8 > [19] 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 > 0.8 0.8 > > ...through to 1567 > > > cor$cor > [1] 0.8 > > boxplot(cor$cor.genes)