Is there any formula which is needed to fix sample size for calculation of allele or genotypic frequency of a certain trait (eg ABO blood group) using Hardy-Weinberg formula? if so for the scenario given by me of the 1000 people how many will be needed as sample?
We always want sample to be representative of the study population. To calculate Hardy-Weinberg formula do we need such representation?
For confidence level 95 % the sample size could be 278 and for confidence level of 99% the sample size could be 400 if the population is of 1000.
Hi,
There is no straightforward answer to your question, as the standard error of a frequency estimate is greater the closer the frequencies are (i.e. if two allele frequency estimates p and q are close to p=q=0.5, the standard error of the estimate will be larger than if p>q or q>p). What I can answer for sure is that the sample does always have to be representative. But to assure representativeness the sampling design is more important than sample size.
Hi all,
It depends on how much "representative" you want your estimation to be. I guess you want to avoid biases and achieve a reasonable precision. For this second point, the statistic of interest is the standard error.
The standard errors of the allele and genotype frequencies depend on the frequencies themselves. Therefore, to achieve a given standard error, you need to make assumptions about the allele frequencies (in addition to Hardy-Weiberg assumptions if you work with genotype frequencies).
There is certainly information around about what would be a good prior guess for these frequencies. Nevertheless, a safe strategy would be to assume that two of the alleles are at a frequency of 0.5 and the third one is absent. In this worst case scenario, the sampling variance of some of the allele and genotypes frequencies would be maximal.
You should have a look at the multinomial (not binomial, because there are three alleles) distribution to understand the sampling process, and how to compute the sampling variances (wikipedia will give you the straightforward formula, from which you can easily get the standard error).
Besides, I agree with Gonzalo that the sampling design is crucial. Think carefully about it in order to avoid biases, i.e. achieve a high accuracy (while the question of sampling size is related to precision).
Cheers
Dear Gourab
The sample size from finite population can be determined by the attached chart :
Good Luck
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size