I beg to differ. I think that there are two extreme types of statistical inference.
1. Comparative inference
2. Representative inference
The former is what is practiced in the fleld of clinical trials and the latter in the field of sample surveys. Clinical trials are really designed to show that treatment can have an effect. The question as to whether they will have an effect is much more difficult to answer but the excuse for ignoring it is that medicine made such a terrible job of answering the first question over centuries that it is worth trying to answer this question well without worrying too much about the second question. The issues are discussed in a paper of mine referenced below.
Randomisation is not about sampling bias. It makes no contribution to this. It is about comparative inference and in that context not just about bias either. It is a method that matches allocation and appropriate inference.
So my take is that randomisation is a means of strengthening the comparative inferences we make and that this is worth doing.
Stephen
Reference
Senn, S. J. (2004), "Added Values: Controversies Concerning Randomization and Additivity in Clinical Trials," Statistics in Medicine, 23, 3729-3753.
Stephen Senn
Professor of Statistics
School of Mathematics and Statistics
Direct line: +44 (0)141 330 5141
Fax: +44 (0)141 330 4814
Private Webpage: http://www.senns.demon.co.uk/home.html
University of Glasgow
15 University Gardens
Glasgow G12 8QW
The University of Glasgow, charity number SC004401
________________________________________
From: Evidence based health (EBH) [[log in to unmask]] On Behalf Of jo kirkpatrick [[log in to unmask]]
Sent: 22 June 2011 15:26
To: [log in to unmask]
Subject: Re: What does randomisation achieve?
I have often wondered about this myself. It is intended to remove sampling bias from a potentially unbalanced population but I think it can be counter-productive. It can just as easily cause the researchers to miss out the people they really should be studying and include those that are a waste of time. When inclusion/exc lusion criteria are used, randomisation is compromised. It is compromised further by the frequent requirement for participants to be from a particular population, eg have an illness or psychological condition or not to be taking certain medications. Then there is the ever present problem of getting a large enough sample for real randomisation, rather than self-selection as participant agreement is always the bottom line. All this brings me to the conclusion that random sampling just pays lip-service to the scientific tradition and in reality perhaps true randomisation, like true researcher objectivity is a bit of a myth. A better option would be strategic sampling, where they choose the best participants for the purpose from the available population.
Best wishes Jo
________________________________
From: Michael Power <[log in to unmask]>
To: [log in to unmask]
Sent: Wed, 22 June, 2011 8:47:45
Subject: What does randomisation achieve?
Nice puff for the EBH discussion list in this interesting article:
What does randomisation achieve?
Adam La Caze, Benjamin Djulbegovic, Stephen Senn
Evid Based Med published 21 June 2011, 10.1136/ebm.2011.100061
http://ebm.bmj.com/content/early/2011/06/20/ebm.2011.100061.extract?papetoc
Michael Power
|