research advice please:
I am administering a drug in mice (n=x), at 4 different dosages, so 4 different group, one of which will act as a control.
After 60 days of drug administation the mice will be sacrificed & a particular enzyme activity will be measured (alpha, beta and game secretase activity, if you're interested). In that 60 day period I will also measure cholesterol distribution (day 0,1,4,10 & every 10 days after). To see what kind of effects the drug has on the things I'm measuring.
So: i have 4 groups with different drug dosages and measure all on different things. What kind of analysis do you recommend, and how many mice do you think I need to still get reliable data
I forgot everything I know about stats, so you'd save me if could just tell what kind of test and the sample size I should use. And perhaps how you calculated the sample size.
Asking this at /sci/ didn't get me far.
>>17142742
Why would we get you any further
Probably you can Google that. Depends on the variation you expect. 3 animals per condition could do. 10 will surely be enough.
>>17142742
How the fucking hell are you allowed to conduct experiments on living organism and "sacrifice" them without even knowing the basics of experimental method?
The general rule is that you should use a sample of at least 5. You can use a smaller one (3 or 4) if you are confident that the drug will work, but then you will have to face harsher criteria for the statistical tests proving its efficacy - and it's the tests the scientific community is looking at, so you'd like to have a high t-value (in case of T-test, here you would probably want to use ANOVA).
Do a power analysis in SPSS, you numb fuck. For most research institutions you can't even get IACUC approval without a power analysis anyway because you have to justify how many animals you are using.
>>17143913
>How the fucking hell are you allowed to conduct experiments on living organism and "sacrifice" them without even knowing the basics of experimental method?
The decline of the West.
It's real you know!