Lecture 10-12 Flashcards
What does the Mann-Whitney U test for?
tests for differences in the CENTRAL TENDENCY of 2 INDEPENDENT samples of ratio-interval scale data when one or both the assumptions are not met
what type of test is the Mann-Whitney U test an alternative for?
2-sample t-test
(Mann-Whitney U test) H0/HA
H0: population1 = population2 || HA: population1 ≠ population2 || does not include μ but compares the central tendencies of both populations
data transformation for Mann-Whitney U test
converting original data into RANKS = ranking transformations – both samples of data are ranked together as one large data
(Mann-Whitney U test) degrees of freedom
sample size (n) || 2 DF needed → n1 & n2 (smallest n, larger n)
outliers
data points that are much more extreme than all of the other data in the sample
how do outliers affect samples?
makes them non normal or heteroscedastic = data rank transformation in order to treat outliers the same as the rest of the data
paired-sample t-test
pair 2 samples via biological link | tests for differences in the means of two paired samples of ratio-interval scale data || does not require variance ratio testing
how are paired-sample t-tests conducted?
work with the differences (subtraction) rather than the actual samples themselves
(paired-sample t-test) H0/HA 2-tailed
H0: μd = 0 | HA: μd ≠ 0 where d = differences
(paired-sample t-test) test statistic
tests one sample of differences being compared to zero, which indicates no difference between the two original samples
(paired-sample t-test) degrees of freedom
sample size of differences = nd - 1 (**need to have equal sample sizes)
(paired-sample t-test) RSNDP
will state if not met
(paired-sample t-test) sample size
must match with each other, number of differences must match n
Wilcoxon paired-sample test
tests for differences in the central tendency of two paired samples of ratio-interval scale data when normality is not met
