Does anyone know of any good software for identifying a minimum number of samples needed to obtain, say for example, 80% power at a significance level of 0.05 for a study? Or would you use power tables found online?

Also, if a piece of software/power table says the minimum number of samples is e.g. n=72? Does that mean you only need to test 72 samples? Or does it mean you need to test more than 72 and need to find the effect you're looking for in a minimum of 72 of those samples? Sorry if this sounds waffly - I'm new to this and although I've read around and asked people in my department, no one seems able to point me in the right direction. If anyone has any pointers or ideas for where to look I'd be very grateful!!

Thanks,
Tulip

I believe most of the major stats packages that you might use for analysing your data e.g. Stata, R will have a function to do this. But nothing wrong with doing a more manual calculation.

If I've understood correctly in your example above it means that you must test at least 72 samples in order to detect any effect at whatever significance level you specified in order to calculate that number, based on the other assumptions you have specified i.e. estimated effect size. Given that these assumptions may be a bit off, and that things can go wrong it is usually preferable to aim for a higher number than specified to allow for this. For example, in a clinical trial, recruiting more people than the minimum is helpful in case some drop out; in a lab study testing extra samples is useful in case an assay fails or has a degree of uncertainty about the result.

Hope this helps

In short, you need to have 3 considerations in mind:

1. What type of test you will use
2. How many participants
3. probability level you will accept as significant (p<0.05 is the most common)

You can do this calculations with G*Power, a free software available online. So based on the above 3 criteria you insert, you should achieve a power of 0.8 at least.

Hi guys,

I've been awful at using this forum lately so I thought I'd already replied to say thank you! Thank you for the very helpful replies, they're much appreciated. I'd never come across G*Power, it's definitely very useful to know about.

Cheers,
Tulip