Dear Yezhou Sun, It seems apparent that your design matrix is actually a character matrix or a data.frame with character or factor entries. Although you say that you converted the design matrix to a numeric matrix, you can't have done so because you would not get the error that you report below. It would help if you told us what commands you used to create the design matrix, and also how you attempted to convert it to a numeric matrix. Then we could tell you how to correct it. Despite several emails to Jim MacDonald, you don't seem to have given this information. Please reply to the mailing list. Best wishes Gordon >[BioC] limma question >Sun, Yezhou yezhou.sun at mssm.edu >Wed Feb 22 23:27:27 CET 2006 >Hi, there, > >I am using limma for some single channel microarray data. Date format >follows that suggested by user guide: probes in row and arrays in >column. Data set has no missing value and was read into R by read.table >function. Design matrix is as following: > > P R >P1 1 0 >P2 1 0 >P3 1 0 >P4 1 0 >P5 1 0 >P6 1 0 >P7 1 0 >R1 0 1 >R2 0 1 >R3 0 1 >R4 0 1 >R5 0 1 >R6 0 1 > >When I ran lmFit(data, design), I got this error: > > fit<-lmFit(data, design) >Error in qr(x) : NA/NaN/Inf in foreign function call (arg 1) >In addition: Warning message: >NAs introduced by coercion > >Traceback gave this: > > traceback() >3: qr(x) >2: nonEstimable(design) >1: lmFit(data, design) > >I have converted data and design to numeric matrices but they never >worked. Could you please what's the problem? I really don't know what's >going on with it. Thanks a lot! > >Yezhou Sun