Table 1.10 - p22

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

Table 1.10 - p22
Someone posted the following message (which reached my inbox but was not actually on the forum as there seems to be some problems with nabble forums at the moment):


If there are 80 people - 40 male, 40 female & half of them are given the drug and half of them given placebo.
Aren't we expecting to have:
1) 20 male given the drug
2) 20 male given placebo
3) 20 female given drug
4) 20 female given placebo?

The table shows:
1) 30 male given the drug
2) 10 male given placebo
3) 10 female given the drug
4) 30 female given the placebo

Can someone please clarify?


The whole point about the example is that it shows what could 'go wrong' if you are unable to control every possible factor in an experiment like this. The idea is that we want to test a drug. So it is reasonable to test it out on a number of people and make sure that half of them get the drug and half do not. Then someone points out that it makes sense to do it on a sample which is representative of the population - i.e. make sure half are women and half are men. But already it becomes difficult to ensure in practice that you can satisfy the 'perfect' design (the 20:20:20:20) option above. And even if you could someone could then say that another factor: "Caucasian or not Caucasian' has to be consdiered. So then we have to ensure that half the people are Caucasians and hal not. But then we have to find a set of people who can be divided (10:10:10:10:10:10:10:10:10) to ensure the perfect balance. Now another factor is ocnsider 'smoker/non smoker' ... you can see that it never stops and you can NEVER design the experiment to control for all possible factors.

Norman Fenton