it is said to be important to determine sample size in any research study, does sample size determination vary between designs and can anyone suggest the easiest way to calculate/determine the sample size for genetic study ? article suggestion is allowed. thanks
Hello,
I read one article related your question few days back. You can go through it.
It was
Sample Size and Statistical Power Calculation in Genetic Association Studies
Eun Pyo Hong and Ji Wan Park
in Genomics and Informatics.
I think you can possibly use GPower to determine adequate sample and effect size for your study
Hello,
You may take a look at this Question/Answers from RG:
https://www.researchgate.net/post/What_is_the_minimum_sample_size_for_population_genetic_study_using_SNPS
I presume that you are referring to an association study for a complex trait. If you are not, skip to the last two paragraphs.
To put it simply, the sample size is determined by the resources you have. Usually, this means your budget. You recruit and genotype as many subjects as you can afford. Usually, the limiting cost is the cost of genotyping. In a rare phenotype, the limitation may be the cost of recruiting. If you are in a hurry, the limitation may be time!
This determines your sample size but you are not done with statistical power considerations. The question has now changed from “what sample size do I need” to “what can I hope to find with the sample size that I can afford.” Easy. I use Shaun Purcel’s Genetic Power Calculator http://pngu.mgh.harvard.edu/~purcell/gpc/ although I am sure that more sophisticated software exists.
There are now thousands of complex trait loci replicated at genome-wide level, so we have a very good idea of what one can expect. Extremely few of loci for discrete traits have a genotype relative risk (RR) higher than 1.5 for variants with >5% minor allele frequency (MAF). So, if your power calculations show that you can only discover loci with RR and/or MAF values higher than that, don’t bother applying for a grant, it will be rejected. This usually means at least 1k cases and 1k controls. For quantitative-trait loci (QTLs), same consideration after you replace RR with the beta value (not sure what the cutoff would be there but look at the QTL literature).
If it is not a complex trait you are after, you are probably looking at a rare disease where the family history suggests monogenic inheritance. There, the answer is the same. Find as many cases as you can, trying your best with international collaborations. Exome them all, ask questions later. The cost has become trivial nowadays, so you don’t need to apply for a grant that will be rejected if you don’t have the required power. If inconclusive, put the data in a public database, hoping someone else has cases that can be combined with yours.
If your question is on population genetics, this is a different discussion on ResearchGate, that an earlier answer has referred to.
Dear Fadaka
Sample size depends on standard deviation ( from previous studies ) and the margin of error you decided .
G*Power software is effective tool to calculate sample size for many ranges of experiments. Also, you can determine effect size and power of the test, G*Power is free to download and easy to use after reading the manual, the download link:
http://download.cnet.com/G-Power/3000-2054_4-10647044.html
The link to download G*Power manual:
http://www.gpower.hhu.de/fileadmin/redaktion/Fakultaeten/Mathematisch-Naturwissenschaftliche_Fakultaet/Psychologie/AAP/gpower/GPowerManual.pdf
Good Luck
Addendum to my answer (see above)
1. I forgot to mention what level of statistical power is acceptable. In a single-hypothesis study, the threshold for acceptable power has been traditionally (and arbitrarily) set at 80%. However, in a high-order multiple hypothesis testing expecting multiple true positives (such as in a GWAS for a complex trait), I would gratefully accept 10-20% or more.
2. The beta threshold for a QTL is not as easy to define as the RR threshold for discrete traits. It depends on the complexity of the QTL, I guess. For a simple eQTL, ( e.g. the blood level of hormone produced by a single gland) one might be allowed to expect higher betas than for a much more complex trait (e.g. a behavoural trait).
Dear RG colleague,
Determining the sample sizes involve resource and statistical issues. Usually, researchers regard 100 participants as the minimum sample size when the population is large. However, In most studies the sample size is determined effectively by two factors: (1) the nature of data analysis proposed and (2) estimated response rate.
For example, if you plan to use a linear regression a sample size of 50+ 8K is required, where K is the number of predictors. Some researchers believes it is desirable to have at least 10 respondents for each item being tested in a factor analysis, Further, up to 300 responses is not unusual for Likert scale development according to other researchers.
Another method of calculating the required sample size is using the Power and Sample size program (www.power-analysis.com).
Regards,
- Badges
- Science topic
- Similar topics
- Composition Studies
- Articles