Thanks for the reply. This is why I posted the above. I actually have a background in pure mathematics and physics at university level before becoming a physiotherapist. And in terms of practical statistical analysis that's the place to understand it. A PHD has no more merit than a school leaving certificate and most researchers employ the services of a mathematician to work out the appropriate stats and power (sometimes) of their research. A few look at it before data collection (perhaps biasing methodology). A few look at it after data collection, again making a statistical analysis fit their needs.

Researches need to under stand the statistical relevence and power of a statistical models results before they even plan their study. This is the first thing to change. Don't do a study to try to make an analysis of when the study doesn't fit. Otherwise the conclusions will be weak at best. It is interesting that when one looks at ratings on the PEDro database that very often they fall below 5/10.

I am a fan of the single case design. Humans are individual. Single case is robust and very powerful. 100 single case designs that show physiotherapy as effective in some way as opposed to not would indeed be a more appropriate look at how we make a difference to the individual. Grouping this into double blind RCT's simply does not work and is a complete waste of time and money.