Dear friends,

​I am a new limma user and keep getting an error code following the topTableF. I have not been able to find any information about the F-statistic not being found in the fit. Any suggestions? I am running Windows 10 and using R. 3.4.3.

​Thanks!

> fit = lmFit((data), mod)
> fitContrasts <- contrasts.fit(fit,contrasts=contrast.matrix)
> eb <- eBayes(fitContrasts)
> results <- topTableF(eb, adjust="fdr",number=length(eb$coefficients))

Error in topTableF(eb, adjust = "fdr", number = length(eb$coefficients)) : 
  F-statistics not found in fit

> sessionInfo()
R version 3.4.3 (2017-11-30)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)

Matrix products: default

locale:
[1] LC_COLLATE=English_United States.1252  LC_CTYPE=English_United States.1252   
[3] LC_MONETARY=English_United States.1252 LC_NUMERIC=C                          
[5] LC_TIME=English_United States.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] bindrcpp_0.2        limma_3.34.9        sva_3.26.0          BiocParallel_1.12.0 genefilter_1.60.0  
 [6] mgcv_1.8-23         nlme_3.1-131.1      forcats_0.3.0       stringr_1.3.0       dplyr_0.7.4        
[11] purrr_0.2.4         readr_1.1.1         tidyr_0.8.0         tibble_1.4.2        ggplot2_2.2.1      
[16] tidyverse_1.2.1    

loaded via a namespace (and not attached):
 [1] Rcpp_0.12.16         lubridate_1.7.2      lattice_0.20-35      digest_0.6.15        assertthat_0.2.0    
 [6] psych_1.7.8          R6_2.2.2             cellranger_1.1.0     plyr_1.8.4           stats4_3.4.3        
[11] RSQLite_2.0          httr_1.3.1           pillar_1.2.1         rlang_0.2.0          lazyeval_0.2.1      
[16] readxl_1.0.0         rstudioapi_0.7       annotate_1.56.2      blob_1.1.0           S4Vectors_0.16.0    
[21] Matrix_1.2-12        splines_3.4.3        foreign_0.8-69       RCurl_1.95-4.10      bit_1.1-12          
[26] munsell_0.4.3        broom_0.4.3          compiler_3.4.3       modelr_0.1.1         pkgconfig_2.0.1     
[31] BiocGenerics_0.24.0  mnormt_1.5-5         matrixStats_0.53.1   IRanges_2.12.0       XML_3.98-1.10       
[36] crayon_1.3.4         bitops_1.0-6         grid_3.4.3           jsonlite_1.5         xtable_1.8-2        
[41] gtable_0.2.0         DBI_0.8              magrittr_1.5         scales_0.5.0         cli_1.0.0           
[46] stringi_1.1.7        reshape2_1.4.3       xml2_1.2.0           tools_3.4.3          bit64_0.9-7         
[51] Biobase_2.38.0       glue_1.2.0           hms_0.4.2            survival_2.41-3      parallel_3.4.3      
[56] yaml_2.1.18          AnnotationDbi_1.40.0 colorspace_1.3-2     rvest_0.3.2          memoise_1.1.0       
[61] bindr_0.1.1          haven_1.1.1  

eBayes doesn't report F-statistics if your design matrix is not of full rank. I daresay that running is.fullrank(mod) would give you FALSE. Remove linear dependencies in your design matrix and try again. In case you're wondering how to do that, here's a little illustrative example:

     Coef1 Coef2 Coef3
[1,]     1     0     1
[2,]     1     0     1
[3,]     1     1     0
[4,]     1     1     0

This matrix is not of full (column) rank, because Coef1 is simply the sum of Coef2 and Coef3. This means that it is impossible to obtain a unique solution for the coefficients, e.g., a solution of Coef1=0, Coef2=1 and Coef3=1 would give the same fitted values as Coef1=1, Coef2=0 and Coef3=0. You need to set up your design matrix to avoid these situations.

Please don't use topTableF(). It's a low-level function intended for internal limma use rather than to be called directly by users. If you look at the example workflows, you will see that topTable() is always used, so just stick to that. The call

results <- topTable(eb, n=Inf)

is equivalent to your use of topTableF.

As Aaron has explained, we think that your design matrix is probably over-parametrized. If that is so, then you would have already got a message from lmFit() to alert you.