Power Analysis for ROC curve

Ask here about academic and research issues, like designing studies, recruiting participants, choosing statistics, submitting for publication, etc.
Post Reply
Johanna89
Posts: 20
Joined: Thu May 09, 2013 11:50 am

Power Analysis for ROC curve

Post by Johanna89 » Mon May 11, 2015 3:44 pm

Hello everyone, I am wondering if anyone has conducted a power analysis for receiver operating characteristic analysis before? I am not quite sure about which system to use when working with categorical outcome data.

If anyone has had a similar experience or has any words of wisdom on this it would be much appreciated!

Many thanks,

Johanna89

Pickle
Posts: 95
Joined: Wed May 07, 2008 6:31 pm

Re: Power Analysis for ROC curve

Post by Pickle » Sun May 24, 2015 5:38 pm

Hi Johanna,

I used MedCalc to do an a priori power analysis (my research relied on ROC Curves) - you can download it for free for 30 days but it's worth seeing if your university have a licence for a full download as you'll need it for the analysis anyway. SPSS will do ROC curve analysis, but MedCalc is superior in the amount of information it generates. E.g. it provides the Youden Index which SPSS doesn't (so you'd have to do this by hand). SPSS bizarrely gave me cut-offs in half numbers which MedCalc didn't.

Metz (1978) recommends suggest 100 cases are required, but I had significantly less and the analysis was fine. Look at similar research to see what other authors reported.

I'd also have a look on Youtube for info about this method - there are some really good tutorials online that explained it better to me in 30 mins than six hours of lectures at uni.

It's a very simple method once you understand what you are doing and why. Good luck!

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest