[BioC] LIMMA: Error in rowMeans(y$exprs, na.rm=TRUE):'x' must be numeric

[James W. MacDonald]

Hi Kripa,

On 10/11/2012 1:37 PM, Kripa R wrote:
> Hi, changing the data to a matrix doesn't help. I think it has 
> something to do with my design matrix which has NAs
>
> design <- model.matrix(~0+group, model.frame(~0+group,data, 
> na.action=NULL))
> > design[55:65,]
>    group1 group2
> 55      0      1
> 56      1      0
> 57      1      0
> 58      0      1
> 59      1      0
> 60      0      1
> 61     NA     NA
> 62      1      0
> 63      0      1
> 64      0      1
> 65      1      0
> So right now dim(design): 247 2 and dim(data) 39000 247
> But doing the next two lines results in the same error
> data<-as.matrix(data)
> fit <- lmFit(data,design)
> /Error in qr.default(x) : NA/NaN/Inf in foreign function call (arg 1)/
>
> How could I tell lmFit that when it reaches an NA in the design it 
> should just ignore the data in the corresponding data column?

You don't. If you don't want to fit a model to all of your data, just 
subset to the data you are interested in and go from there.

Best,

Jim



>
> Thanks,
>
> .kripa
>
> > Date: Thu, 4 Oct 2012 13:05:33 -0400
> > From: jmacdon at uw.edu
> > To: kripa777 at hotmail.com
> > CC: bioconductor at r-project.org
> > Subject: Re: [BioC] LIMMA: Error in rowMeans(y$exprs, 
> na.rm=TRUE):'x' must be numeric
> >
> > Hi Kripa,
> >
> > On 10/4/2012 12:54 PM, Kripa R wrote:
> > > Hi so I'm trying to do limma fitting on my data but i keep getting 
> the following error
> > > Error in rowMeans(y$exprs, na.rm=TRUE):'x' must be numeric
> > > But I'm not sure what the problem is does anyone have suggestions?
> > >
> > > Here's what i've been doing after normalization:
> > > data<-subset(data_pheno, select=is.na(data_pheno['case',])==FALSE) 
> #remove columns with NA
> > > group<- factor(data['case',])
> > > design<- model.matrix(~0+group) #dim 240 2
> > > expr<-as.data.frame(data[1:39693,]) #only keep rows that need to 
> fit aka expression data, dim 39693 240
> > >
> > > fit<- lmFit(expr,design)
> > > Error in rowMeans(y$exprs, na.rm=TRUE):'x' must be numeric
> >
> > The help page for lmFit() lists lots of possible input object types, 
> but
> > data.frame is not one of them. Your code is fairly cryptic, so I 
> have no
> > idea what is in that data.frame, other than the obvious conclusion that
> > there are some non-numeric data.
> >
> > You would be better served to use existing infrastructure for 
> processing
> > your data rather than a data.frame. I think there should be enough
> > examples in the limma User's Guide to show you the way.
> >
> >
> > Best,
> >
> > Jim
> >
> >
> > >
> > >
> > > Thanks,
> > > .kripa
> > > [[alternative HTML version deleted]]
> > >
> > > _______________________________________________
> > > Bioconductor mailing list
> > > Bioconductor at r-project.org
> > > https://stat.ethz.ch/mailman/listinfo/bioconductor
> > > Search the archives: 
> http://news.gmane.org/gmane.science.biology.informatics.conductor
> >
> > --
> > James W. MacDonald, M.S.
> > Biostatistician
> > University of Washington
> > Environmental and Occupational Health Sciences
> > 4225 Roosevelt Way NE, # 100
> > Seattle WA 98105-6099
> >

-- 
James W. MacDonald, M.S.
Biostatistician
University of Washington
Environmental and Occupational Health Sciences
4225 Roosevelt Way NE, # 100
Seattle WA 98105-6099