S2 lecture 5 - between-subject designs

Question 1

What is t distribution?

Answer 1

the standardised mean-difference distribution. Can be calculated for any mean difference. The score is then compared to a table to check for statistical significance.

Question 2

How do we inturpret t?

Answer 2

the value can be positive or negative depending on the order in which you subtract the means from each other (so keep track of this). The larger the value of t (whether + or -), the stronger the evidence against the null hypothesis.

Question 3

If there is a large distance between means and not much variance within each condition, the value of t will be…

Answer 3

high

Question 4

If there is a small distance between means and a large amount of variance within each condition, the value of t will be…

Answer 4

low

Question 5

What are the two main t-tests?

Answer 5

independent-sample t-test: for between-subject design
paired-sample t-test: for repeated-measure design
there is also the one-sample t-test: for comparing one mean to a specific number, but this is not very common.

Question 6

Why should you avoid saying “the two groups had different scores, but the difference was not significant” with regards to statistical significance?

Answer 6

No statistical significance means it’s not appropriate to conclude there was a difference at all.

Question 7

What are the assumptions that must be met for an independent t-test?

Answer 7

Variances of each group should be approximately equal, the DV must be interval or ratio, each data point must be independent, sampling distribution must be normally distributed.

Question 8

Why must data be ratio or interval for independent t-test?

Answer 8

Because of the mathematics of t distribution, the DV must be continuous, where intermediate values (1.24,3.5, etc.) make sense, versus discrete values which are essentially categories, and fractional values would not make sense.

Question 9

How do you check for normality of sampling distribution?

Answer 9

examine the histogram of your sample for obvious signs of non-normality (such as skewness, 2 peaks), look at skewness and kurtosis from explore output, use a K-S test.

Question 10

Is a t-test parametric or nonparametric? What does this mean?

Answer 10

A t-test is parametric, meaning it makes certain assumptions about distributions that are required for the test to work mathematically. Nonparametric tests don’t make these assumptions, but they tend to be less powerful.

Question 11

What are two nonparametric tests you can use if t-test assumptions are violated?

Answer 11

Mann-Whitney U Test for between-participant designs, Wilcoxon signed-rank test for within-participant designs.