Unless I'm missing something, there seems to be a bug in the variance stabilization function vsn2 or the predict method for the vsn object when calib="none". Below is a self-contained example with random data. Thanks!

>#Simulate data
>library(vsn);
> nGenes = 1000;
> nSamples = 100;
> data = exp(matrix(rnorm(nGenes * nSamples), nGenes, nSamples) + log(1:nGenes));
### Run vsn2 with default arguments: no problems
> fit = vsn2(data)
vsn2: 1000 x 100 matrix (1 stratum). Please use 'meanSdPlot' to verify the fit.
> vsdata = predict(fit, newdata = data);

### Run vsn2 without calibration
> fit2 = vsn2(data, calib = "none")
vsn2: 1000 x 100 matrix (1 stratum). Please use 'meanSdPlot' to verify the fit.
> vsdata2 = predict(fit2, newdata = data);
Error in vsn2trsf(newdata, object@coefficients, strata, hoffset = if (log2scale) object@hoffset else NULL) :
  'p' has wrong dimensions.

> sessionInfo()
R version 3.3.1 Patched (2016-07-30 r71015)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Fedora 24 (Twenty Four)

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

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

other attached packages:
[1] vsn_3.42.3          Biobase_2.32.0      BiocGenerics_0.18.0

loaded via a namespace (and not attached):
 [1] Rcpp_0.12.6           lattice_0.20-33       grid_3.3.1           
 [4] plyr_1.8.4            gtable_0.2.0          affy_1.50.0          
 [7] scales_0.4.0          ggplot2_2.1.0         BiocInstaller_1.22.3
[10] zlibbioc_1.18.0       affyio_1.42.0         limma_3.28.17        
[13] preprocessCore_1.34.0 munsell_0.4.3         colorspace_1.2-6     

Dear Peter

Thank you for pointing this out and the clear bug report. This should be fixed in the devel branch in vsn >= 3.43.9, which is in the subversion repository now and should be online in a couple of days.

As the vsn2 function's manual page says, the calib="none" option is an experimental feature, and even though it was entered more than eight years ago, I am not sure it was ever properly tested, and certainly I have not had my mind on it for a while. The bug that you observed (and that I now fixed) is that this option was not passed down to a helper function within the package. In any case, I would be interested in your mileage with this functionality and any feedback you may have.

Kind regards

Wolfgang