PaGE 5.0 Documentation - the "log transform" option

You can input your data in log scale, or not. When you select File New Analysis Type, you will be asked if your data is logged or not.

When using the t-statistic you can apply it either to the unlogged data, or to the logged data. This is true regardless of number of channels or the design. If your data is logged and you choose to run it on the unlogged data, the program will unlog it for you.

When you choose Run PaGE by t-stat, you will be asked whether you want to run on the Logged or the Unlogged data.

Note 1: Using the logged versus the unlogged data can make a difference to the results. Unfortunately there is no general indication of which is preferrable, and in fact each might find genes the other does not (see the technical manual for examples). So this is a parameter that you might want to try both ways and see which gives more satisfactory results. Note that the confidence reported is valid in either case, what can be affected is the power of the test.

Note 2: If you use this argument and your input data is unlogged, but contains non-positive values, then the program will complain. You can proceed only if you shift the data to be all positive, and the program will suggest a moderate shift to accomplish this.

Shifting data, however, has subtle consequences to the results, so should be done with care.
The only kind of data we are aware of that currently produces unlogged negative intensities is that output for Affymetrix arrays by Probe Profiler, or from any of the Affymetrix MAS 4.0 algorithms or earlier.