Question
Asked 11 May 2014

What are other ways of calculating estimated sample size besides using effect size?

I will be doing a control group of older people (without intervention) and was thinking of using its effect size to estimate the sample size of my intervention group further down the timeline. But won't the effect size of the control group (without intervention) be smaller and hence, potentially end up with a bigger than needed intervention group sample size estimate? I am not familiar with statistics and this is my rationale only. Would appreciate your suggestions and other considerations.

Most recent answer

All Answers (4)

Jochen Wilhelm
University of Giessen
The effect you are actually interested in is the *difference* in response between your intervention group and your control group. It looks like a two-factorial design (one factor is the "time" (t0, t1, maybe more time points, maybe categorical or metric) and the other factor is "intervention" (binary: yes/no). The specific intervention effect is termed "interaction" of the two factors. For this you should define a relevant size and calculate the required sample size.
Andrew Ekstrom
Oakland University
If you are only looking at 2 groups and only care about the effect of intervention (add don't care about all the other important factors) you will use something like a t-test. Something I found interesting for dividing up the total number of participants, 'N' is to use the equation, n1 = N*S1/(S1+S2). If you know the standard deviations of each group, S1 and S2, the formula above will give you the most power in your test. I would suggest that you take all the info you can get from each person and use it in your analysis.

Similar questions and discussions

For the scatterplot at the attached link, what are the most important considerations for a regression evaluation?
Question
21 answers
  • James R KnaubJames R Knaub
This is a question for philosophical discussion: I saw the interesting example at the link below (excerpt from East Tennessee State University on Freedman, et.al.). It is meant to be a simple example, and only has 11 data points, so even if other regressors were available, we can only speculate about more data points, but it appears that there could be a number of phenomena that might be at work: (1) leverage, (2) stratification, (3) heteroscedasticity, (4) multiple regression, (5) nonlinear regression. A confidence bound on the slopes that could occur with OLS would be wide at each end. However, it appears that it might go through the origin, which would make confidence on a slope appear as two straight lines, forming a v-shaped wedge. But, there are people who die of lung cancer without smoking, so one would expect to see a positive intercept. Regardless, larger x would generally mean larger residuals for y.
What are your thoughts on this graph? What would you tell people who might ask you to evaluate what may be (or have been) happening?

Related Publications

Article
Full-text available
Abstrakt: Aplikovaný štatistický výskum zohráva kľúčovú úlohu pri riešení problémov v rôznych oblastiach ekonomických, spoločenských, technických, poľnohospodárskych vied ale aj zdravotníctva. Mnohé zisťovania sa vykonávajú výberovým spôsobom. Technika výberu vzorky a určenie veľkosti vzorky majú rozhodujúcu úlohu pri záveroch o populácii na základ...
Got a technical question?
Get high-quality answers from experts.