issues with NHST logic / one-sample t-tests

main problem with NHST

only produces a binary decision

- lacks important context; does not indicate magnitude of difference between x̅ and μ

effect size (Cohen's d)

magnitude of difference between what we observed in our sample and what we expected from the population

- estimate of the standardized difference between x̅ and μ

(.2 or .3 = small; .5 = medium; .8 = large)

confidence interval

range of values within which a population parameter is estimated to lie (interval estimate)

- tells us about precision of our estimate (the narrower, the more precise)

- only constructed for two-tailed tests

power

indicates how sensitive our test is to rejecting the null hypothesis

(e.g., if 1 - β = 1, there is no chance of a type II error)

type 1 error (false positive)

rejection of a true null hypothesis; alpha

type 2 error (false negative)

failing to reject a false null hypothesis; beta

research design

refers to how research is conducted; broad term to reflect all diff decisions made along the process

of all the things that affect power, which factor can we control as researchers?

the sample size

t-metric

uses an estimate of population variance (standard deviation), rather than population variance itself

- applied to a family of theoretical distributions

degrees of freedom (N - 1)

number of independent pieces of info remaining after estimating one or more parameters

- reduces the bias of our estimation of variance