The limma model is not correct; see the updated paper at Colin's site for a correction.
David Airey, NA
5 September 2014
The design in the original paper repeatedly samples 6 flasks over 5 time points, rather than 30 flasks grown to different time points where each flask is sampled once. The latter design might be appropriately analyzed by the limma code presented in the paper, whereas a better comparison for the actual design with the "timecourse" package would be to consider hierarchical designs in limma. The design of the actual experiment is a simple split plot, with the one between subject (flask) factor being strain and the one within subject (flask) factor being time. The fix here is to note the correspondence between the example in section 9.7 of the limma manual. Substitute time for tissue, and flask for subject, and strain for condition and proceed as shown.
The limma model is not correct; see the updated paper at Colin's site for a correction.
5 September 2014
The design in the original paper repeatedly samples 6 flasks over 5 time points, rather than 30 flasks grown to different time points where each flask is sampled once. The latter design might be appropriately analyzed by the limma code presented in the paper, whereas a better comparison for the actual design with the "timecourse" package would be to consider hierarchical designs in limma. The design of the actual experiment is a simple split plot, with the one between subject (flask) factor being strain and the one within subject (flask) factor being time. The fix here is to note the correspondence between the example in section 9.7 of the limma manual. Substitute time for tissue, and flask for subject, and strain for condition and proceed as shown.
Competing interests
None.