I have data of a cell migration assay that I need to evaluate. I have a number of replicates for each condition and the total number (#) of migrated cells for each replicate and condition (control vs treatment with different substances). Now, there are various ways to do the evaluation - I could do the following comparisons:

* total # of migrated cells under the different conditions
* % of migrated cells relative to total # of seeded cells
* % increase in # of migrated cells vs control mean
* fold change of # of migrated cells vs control mean

Which type of evaluation would you favor and why?

Different measures can be used to highlight different things, typically you select the measure based on what you want to show. For control vs treatment studies often relative difference of some sort is shown (either percentage or fold). Just showing raw counts is usually requires the reader to do extra work to interpret the data, you generally want to avoid this.

Thanks for your answer. I completely agree that selecting the measure that makes it easiest for the reader to understand what I am trying to show is the best option.

One thing that baffles me a bit, however, is that depending on which type of comparisons I chose I get different results in the statistical analysis. So, I might get statistically significant differences in some cases, but not in others. The question remains the same in all cases, i.e. "Are there (most likely) true differences between the various conditions?", but getting different results from the statistical analysis depending on which types of data I use (raw data or converted data) is confusing.

That depends a bit on the nature of the data and I am not entirely sure whether there is an universal answer to that. If migration events are rare, even in the reference sample, for instance your data will be dominated by non-migration events. If you only count the migrated cells and ignore the non-migration events, you would amplify the signal. However, it would also amplify potential skewedness in your data, for example. For the most part I would go for percentages as in many assays I would expect a proportionate increase in migration after treatment, but again, that depends a bit on the assay.