SINGLE SAMPLE T-TEST
Single sample t-test - parametric procedure for testing the Ho for a single sample experiment when the SD of the population must be estimated
Assumptions1. Interval or ratio level scores
2. Normal distribution of population
3. Use sx to estimate _____
*** A robust statistic particularly if N ³ 30
Calculating the single-sample t-test
1. Compute an estimated SD
2. Compute estimated standard error of mean
3. Compute t obt
t distribution - distribution of all possible values of t computed from random samples from the raw score population
*** There are many versions of the t-dist.(different shapes) depending on the size of N
*** Differently shaped t-distributions produce different critical values
Degrees of freedom
*** df = N-1 (this actually dictates the shape of t-distribution)
*** The larger the df the closer the t-dist. is to the normal distribution
*** df > 120 produces a normal curve
*** df > 30 approximates a normal curve
Estimating ____ by computing a confidence interval
Two ways to estimate __________
1. Point estimation
*** ____= ______
*** Vulnerable to sampling error
2. Interval estimation
*** Takes into account the possibility of sampling error
*** Use a margin of error (±) &endash; create an interval (Confidence interval )
*** Compute the highest & lowest value of m that are not significantly different from the sample
mean
59.72 ___________ 71.62 (95% CI)
*** Use 2-tail critical value for CI
*** The smaller the ____, the smaller the probability of an error, so the greater the confidence
Maximizing the power of the t-test
Power &endash; probability of not committing a Type II error
*** Maximize power by maximizing probability of obtaining significant results
3 methods for increasing power
1. Greater the difference produced by changing IV, greater the power
2. Smaller the variability of raw scores, greater the power (minimize within error)
3. Greater the N, greater the power (N = 30 produces minimal power)
*** Designing a powerful study results in a larger OV relative to CV