Hi. Of course, power and sample calculation are two different things. You calculate sample size to know how large the sample should be (E.G. how many patients to include in the trial) to obtain the expected effect with an established power. On the other hand, power calculation is made to establish the minimal effect which can be obtained with a given sample size.
Oct 16, 2012
Sai Kishore
·
Chettinad Hospital & Research Institute
Says:-
Thanks for your nice comment Mr. Lenarczyk. I am keen to know how much of minimal effect can be obtained and how it benefits the research?
That is a decision you have to make. Large samples produce greater effect sizes, but are expensive and time consuming to put together. Does your research need to meet a standard of significance at .05, .01, or greater? That would depend on the standards and expectation of your field of study. If you can't claim positive results at greater than .01, then you need to figure out how many subjects you need to achieve results at that level.
Oct 24, 2012
Sai Kishore
·
Chettinad Hospital & Research Institute
Says:-
Thanks Ms. Bombach. Yes, I do agree that large samples provide better significant result. In order to achieve the same with a smaller sample, (as per the last sentence of your valuable comment) do you suggest a pilot study first? And based on which the sample size can be sorted out and then the rest.
People often do a pilot study so they have a feel for the kinds of outcomes they may get. It also helps in the final research design--a preliminary study is the point to find out if your measures and protocols work before you run a big, expensive study. You can use a pilot to assess the discriminatory power of your treatment levels or conditions in advance. You may want to do a pilot to see if your measurements can pick up on significant but rare outcomes. Some outcomes can be hugely important but have such a relatively low incidence rate that they are not picked up in smaller studies. (Take a look at what happened with Vioxx as an example.)
There are multiple considerations that go into designing a study; some are based on statistical measures of effect size, sample size, and power, while others are based on expense, difficulty of obtaining samples/subjects, or importance of potential outcomes.
Oct 29, 2012
Sai Kishore
·
Chettinad Hospital & Research Institute
Says:-
I understand. Thanks once again Ms. Kathleen. Now this makes me keen enough to ask for more. Calculating the power of study should have a method to calculate. Does this method applies to all the studies, irrespective of whether it is a descriptive or a meta-analysis or cross sectional?
If so, then does doing a pilot also require its power to be calculated?
Power is nothing but probability of not committing Type II error. You can increase the Power of your study by inflating the sample size. As the sample is closer to the population the power also will increase. But as a reaseracher we are interested in achieving this power with an adequate sample size. This can be scientifically calculated based on study designs, statistical tests used etc., with help of formula (conventionally). Nowadays, we have softwares to calculate the power and sample size.
Nov 18, 2012
Sai Kishore
·
Chettinad Hospital & Research Institute
All Answers (8)