EXPERIMENTAL AND QUASI-EXPERIMENTAL STUDY DESIGNS

(TAKEN FROM CAMPBELL & STANLEY)

Beware of the simple two-point result, especially the tendency to

imagine a clear 'improvent' between them

In this model, there seems to be no effect measureable apart from the

long-standing trend....

Similarly here: the intervention appears to be ineffective

Here again, the 'jump' has been lost in the background 'noise'

This could be a real effect and should be studied further

Here also...

And here

This could be a delayed effect.  Worth further study

The advantages of the cross-over trial are that fewer numbers

are needed because ALL people are exposed at different times

to BOTH arms of the study.  Also, the "order" effect can be studied. 

But it demands a full "washout" period to be sure that no residual

remains.  Full double-blinding is essential.