Next: 2.3.2.3 Approximate test with Up: 2.3.2 Examples Previous: 2.3.2.1 Voter preference

2.3.2.2 Deployment of defibrillators

We went through the example of analyzing a study regarding deployment of defibrillators.
• Model: Given that you need resuscitation, it is equally likely to be treated by either methods (CPR or CPR+defibrillator).
• : Equally likely to survive with CPR or with CPR + defibrillator.
• : More likely to survive with CPR + defibrillator.
• : # out of survivors who were treated by CPR + defibrillator. Under the model and the null hypothesis we will have that has a Binomial distribution with parameters and . In other words if you sample a survivor, it is equally likely that it was treated by CPR only or by CPR + defibrillator.
• : .

Remark What if it is more likely for some reason to be treated by CPR+defibrillator rather than by CPR alone? For example say that at most sport centers patients always ended being treated by CPR + defibrillator. This would be a big trouble. We might reject the null hypothesis in favour of when in fact we are really rejecting the model (specifically, that it is equally likely to be treated by either CPR or CPR + defibrillator).

Next: 2.3.2.3 Approximate test with Up: 2.3.2 Examples Previous: 2.3.2.1 Voter preference
Eran Borenstein
2007-05-04