Statistical power in the context of Genome-Wide Association Studies (Coursework Sample)

Name
Instructor
Course
3 April 2014
Epigenetics
Relationship between sample size, statistical power and type 1 and type 2 errors
Sample size has close relations between statistical power and type 1 and 2 errors. Statistical power has a direct proportion to sample size (Houle, Donald, & Chris, 3). Increase in sample size leads to an increase in power. It is therefore noted that any irrelevant but observable difference in clinical study, can become statistically significant with an increase in sample size (Houle, Donald, & Chris, 3; Norton, & Michael, 2). Another way to look at statistical power according to Norton et al. (2) is probability of not making type 2 errors thereby creating a close relationship is created between statistical power and type 2 errors. Statistical power comes in to measure the chance of a test to accurately reject a false null hypothesis (Norton, & Michael, 2). Type 1 error is brought about by a researcher concluding that there is actually a difference while in reality the difference does not exists at all (Norton et al, 2).. Type 2 errors occur when a researcher falsely concludes that a difference does not exist while in reality a difference exists (Norton et al, 2). Both type 1 and 2 errors occur due to the fact that the researcher is not sure of a difference existing between measured means (Evan & Purcell, 665).
Statistical power in the context of Genome-Wide Association Studies and ways to improve power
Statistical power in context of Genome-Wide Association Studies' factors in: (a) effect size (b) sample size (c) correlation with the causal variant and (d) marker allele frequency (Stranger, Barbara, Eli, & Towfique, 373). Since Genome-Wide Association Studies' are unable to identify associations of smaller sizes there is a requirement to provide a larger population sample in order to obtain moderate effects from a small variant of a trait due to Genome-Wide Association. In the process of maximising the Genome-Wide Association Studies' in a studied trait, increasing statistical power and sample size becomes fundamental (Evan & Purcell, 665). In Genome-Wide Association Studies, effects of a trait are only realized by increasing the sample size to reach a higher statistical power (Spencer et al, 2). The value of DNA pooling is a measure towards increasing the statistical power of Genome-Wide Association Studies (McCarthy et al, 359).
Type I error in Genome-Wide Association Studies and common procedures used to deal with it
Type I error is the rejection of a null hypothesis instead of rejecting the researchers hypothesis. In the Genome-Wide Association Studies, researchers may fail to conclude an existence of difference in effects on a trait phenotypically due to smaller sample size of genotype (Bush & Moore, 6). According to Balding, Type 1 error in Genome-Wide Association Studies can be inflated by phasing cases in addition to imputing missing genotypes and controlling selection before testing (783). In order to increase the probability of statistical tests in Genome-Wide Association Studies reporting correct significant results of phenotype, the value for rejecting a valid null hypothesis has to be adjusted to low values.