Hello everyone.
I would like to ask your help on the decision about the best methodology to determine the sample size to conduct an observational study.
This is a study to be carried out by a pharmaceutical company in which it is intended to characterize all combinations of drugs used before, during and after the administration of the drug of interest.
The patients observed are cancer patients with advanced stages of the disease and therefore frequently change drugs.
There is a great possibility of drug combinations, since the order in which they are administered can also change. So, what is intended after the sample is collected is to build a Sankey Diagram.
The primary objective of the study is to characterize the path taken in terms of drugs. From what I understand, this diagram has not involved any statistical tests, so we are talking about making a simple characterization.
It seems to me that each combination of drugs may have a low relative frequency, so I don't know which frequency to use when calculating the sample size.
Is it okay to consider the following calculator?
http://www.raosoft.com/samplesize.html
It is usual to use the frequency 0.5 when we do not know for sure the prevalence we intend to estimate, but here it seems to me a very high value... I think a value close to 10% or 5% would be more correct, but I have no idea and I have fear of making a serious mistake, since I'm not familiar yet with Sankey Diagram.
Can someone help me to clarify this doubt. Thank you very much in advance for your attention.
Adriana
Dear Adriana,
Power analysis is typically used to calculate sample size. The choice of a specific type of power analysis depends on which statistical criterion you will use in the future, in particular, for example, for the analysis of qualitative features, Chi2 test is used with or without Yates' correction, Fisher's test, etc.
For statistical analysis, specialized programs are best suited: SPSS / STATISTICA, or statistical programming languages: SAS / R.
Here is a link to a manual for power analysis in R language: https://www.statmethods.net/stats/power.html
I hope I could partially help you.
- Badges
- Science topic
- Similar topics
- Chemistry
- Pharmacology
- Pharmaceuticals
- Drugs