I have a sample size of 10 participants and get Cohen's d of 1 to 1,3. I would like to know if I can use the online conversion effect size calculator to convert these d in NTT
Typo for NNT (Number needed to treat)?
If so there is no simple calculation as you'd need a threshold for the size of effect that is considered successful treatment. One you decide a threshold you could in principle calculate NNT based on the proportion of participants who cross that threshold. There would be some subjectivity involved in that.
Also with 10 participants the precision of this estimate will be terrible and not tell you much. Also there are critiques of NNT in the literature that also focus on precision among other things so I'm not sure it will be very useful.
Finally I'm not sure why d is varying ... but d of 1+ is quite large in many contexts and might imply all participants will cross the threshold in which case NNT will be 1 and the interval difficult to calculate except by adding information such as a Bayesian prior.
To get an idea of a Bayesian approach (or a crude approximation thereof) using a common prior I get an interval estimate of:
[2.0, 34.0] for a 95% probability interval
This gives an idea of the precision - though with 10 cases the prior is doing a lot of work (i.e., the NNT estimate is around 6) so I'm not sure I'd be comfortable using this except to illustrate or sense check that even with making some further assumptions we aren't getting a very precise estimate of NNT. If you had an informative prior (from say another larger study) that might even be useful a approach ... maybe.
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication