Power and Sample Size
Embed Size (px)
Transcript of Power and Sample Size
Power and Sample SizeBoulder 2006Benjamin Neale
To Be AccomplishedIntroduce concept of power via correlation coefficient () exampleIdentify relevant factors contributing to powerPractical:Empirical power analysis for univariate twin model (simulation)How to use mx for power
Simple exampleInvestigate the linear relationship (r)between two random variables X and Y: r=0 vs. r0 (correlation coefficient).
draw a sample, measure X,Y calculate the measure of association r (Pearson product moment corr. coeff.) test whether r 0.
How to Test r 0assumed the data are normally distributeddefined a null-hypothesis (r = 0)chosen a level (usually .05)utilized the (null) distribution of the test statistic associated with r=0t=r [(N-2)/(1-r2)]
How to Test r 0Sample N=40r=.303, t=1.867, df=38, p=.06 =.05As p > , we fail to reject r = 0
have we drawn the correct conclusion?
= type I error rate probability of deciding r 0(while in truth r=0) a is often chosen to equal .05...why?
N=40, r=0, nrep=1000 central t(38), a=0.05 (critical value 2.04)
Observed non-null distribution (r=.2) and null distribution
In 23% of tests of r=0, |t|>2.024 (a=0.05), and thus draw the correct conclusion that of rejecting r = 0.
The probability of rejecting the null-hypothesis (r=0) correctly is 1-b, or the power, when a true effect exists
Hypothesis TestingCorrelation Coefficient hypotheses:ho (null hypothesis) is =0 ha (alternative hypothesis) is 0Two-sided test, where > 0 or < 0 are one-sidedNull hypothesis usually assumes no effectAlternative hypothesis is the idea being tested
Summary of Possible ResultsH-0 trueH-0 falseaccept H-01-abreject H-0 a1-b
a=type 1 error rateb=type 2 error rate1-b=statistical power
Rejection of H0Non-rejection of H0H0 trueHA true
PowerThe probability of rejection of a false null-hypothesis depends on: the significance criterion ()the sample size (N) the effect size ()The probability of detecting a given effect size in a population from a sample of size N, using significance criterion
Talpha 0.05Sampling distribution if HA were trueSampling distribution if H0 were truePOWER: 1 - Standard CaseNon-centrality parameter
Talpha 0.05Sampling distribution if HA were trueSampling distribution if H0 were truePOWER: 1 - Increased effect sizeNon-centrality parameter
Impact of more conservative Talpha 0.01Sampling distribution if HA were trueSampling distribution if H0 were truePOWER: 1 - Non-centrality parameter
Impact of less conservative Talpha 0.10Sampling distribution if HA were trueSampling distribution if H0 were truePOWER: 1 - Non-centrality parameter
Talpha 0.05Sampling distribution if HA were trueSampling distribution if H0 were truePOWER: 1 - Increased sample sizeNon-centrality parameter
Effects on Power RecapLarger Effect SizeLarger Sample SizeAlpha Level shifts Type of Data:Binary, Ordinal, ContinuousMultivariate analysisEmpirical significance/permutation
When To Do Power Calculations?Generally study planning stages of studyOccasionally with negative resultNo need if significance is achievedComputed to determine chances of success