Data Analysis week 5

winniesmith2's version from 2017-10-30 10:48


Question Answer
What are type 1 errorsalpha error. This is the error level we set when testing the null hypothesis
What happens if you choose a sample that is too small?-Inconclusive results – type 2 errors -Calculations cannot be completed e.g. ,some tests require a certain n -Sample may be unrepresentative
what are type 2 errorsfalse negatives
What happens if a sample size is too large?-Collecting more data than is necessary is costly in many ways -You have the statistical power to pick up smaller and smaller differences and label them as ‘significant’ . Does the difference matter? may be statistically different but is this biological/clinically significant.
An appropriate sample size depends on what-The size of the effect that we think is interesting -The alpha level -The variability we expect to find in our sample
examples page 7 onwards
What is an appropriate sample size?To estimate this we need to know an estimate of the variability we expect in our sample
How to calculate a sample size Specify a sample size to detect a pre-defined level of difference in a statistical test (example used here is t-test). -page 13. EQUATIONS.
what does LSD meanlowest significant difference.
some considerations for working out sample size The difference you want to detect will be based on your knowledge of the variables you are testing from past research. Information on the likely standard deviation needs to come from previous studies. If these do not exist you need to do a pilot study first to gain this information. (these calculations only consider alpha, beta beyond scope of module).
examples use power point
What would you do for a one tailed test change the critical value to reflect a one tailed test . this would reduce the required sample size.