One medicine but three different combinations. Effect size is unknown. The previous data have several dependent and independent variables.
Any answer to this question requires a formal sample size estimation.
To do so, you must know (to DIY) or tell (if you want others to do it) what follows:
- EXACTLY how you will make the comparisons, i.e. how the study is designed
- some previous data allowing to estimate what you expect to observe (mean, SD, correlations if the measures are repeated within subjects...) and how strong the effect could be - or, at least, how strong the effect should be to be relevant for your goal.
- the success rate of the experiment (for instance: if you need to sacrifice the animal and get a tissue for a molecular test, and you can fail 20% of times either in getting the issue prepared or getting the result from the test... then you have to increase the estimated sample size)
HTH!
Marco
Actually all the three treated groups will be given a medication for 2, 5, and 7 days respectively and consecutively. Then they will be challenged with an infection. Then all animal will be sacrificed and blood parameters and histo-pathological examinations will be carried out. This is the experimental design.
There is still a lot of stuff which remains unclear.
For example:
- will the time between the last administration of the medication and the exposition to the infectious agent the same for all the groups? will the daily dose be the same for all the groups (i.e. will the time under treatment the only change between the groups?)
- which parameters will you measure? how do you expect them to be (in terms of mean values and SD? usually, are those variables normally distributed? how strong do you expect / want the differences between groups to be?
- if you have more variables, are them equally important for your study? or just one / a few of them?
- would you like to compare across all groups simultaneously?
So, you have to estimate sample size for every endpoint, making assumptions on its mean/sd and on the difference between the two groups you're comparing each time; for the sake of the estimation, you'd assume you'll analyse data by using a Student's t test for independent groups (or, if the resulting N is less than 25 per each group, you'll need to re-estimate for Mann-Whitney's U test because you wouldn't be able to check for t test's assumptions).
Then, if you can measure all the variables on the same samples (i.e. per each sacrified animal, you test every blood and histo-patho parameter) you can take the highest estimated N.
For the estimation, you can choose among a number of softwares providing tools for sample size & power, like G*Power, Power&Precision, Stata, or packages in R.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size