Flashcards
statistical tests
Cards in this deck (18)
- 1does the observed value have to be greater, the same smaller than the critical value in order to reject the null hypothesisobserved value must be equal to or SMALLER THAN CRITICAL VALUE in order to reject the null hypothesis
- 2what value does cramer's v have to be in order to accept a strong contingencemore than 0.5 is strong contingence
less than 0.5 is weak contingence - 3conditions for chi-squared: No more than ___ % of the categories should have an expected frequency of less than __20%
5 - 4what test would you apply when you have when you have counts in categories arranged as a contingency tablechi squared
- 5what does increasing the sample size do to probabilityincreasing sample size reduces the probability that differences occurred just by luck
(p value becomes smaller) - 6what does increasing sample size do to cramers vcramers v is unaffected
- 7what test would you use when you have a single sample with counts in two categories. The null hypothesis is usually that there is no preference in the population The probability value (p) tells you the chances that the split of answers could have arisen by chancethe binomial test
- 8this is the general hypothesis for which test: Tests the hypothesis that the groups in the columns are contingent upon the groups in the rows.chi squared
- 9what two sample test would you apply when you have independent measurements and want to know if the values in one sample are significantly higher or lower than values in the other samplemann-whitney U-test
- 10when would you use student's t-testtwo samples
independent measurements (not paired)
when sample size is large >30
data needs to be normally distributed - 11what tests would you apply to paired dataWillcoxon's matched-pairs signed-ranks test
or
the sign test - 12what is the difference between Willcoxon's matched-pairs signed-ranks test and the sign testWilcoxon’s test can show how large the difference is, whereas the sign test cannot use this information – Wilcoxon’s is a more powerful test
- 13what test would you apply when: • Three or more categories • Data is linked • nominal categories Null hypothesis: no difference between the groups so we expect the ranks to be randomly distributed between the groups so the sum of ranks is fairly similarfreidmans
- 14what test would you apply when you have: more than two samples independent measurements samples are linked ordinal categoriespage's trend test
- 15which test would you apply when you have: more than two samples individual measurements independent samples nominal categoriesKruskal-wallice H test
- 16k tests tell you if there is a difference. what further step can you take to find out where the differences between groups liePost-hoc testing
- 17Imagine you are in charge of the raffle tickets and you do not want to give away more than 5% winning tickets. everyone gets three dips in the bag. what procedure would you have to carry out in order to keep the 5% limitThe Bonferroni correction procedure
(5/3 = 1.7%) - 18what test would you apply when you have: two samples each sample has counts in categories the categories are ordinalKolmogorov-Smirnov two-sample test.
when the categories are nominal use chi-squared two sample test
Information
- Description
- GEG 100 Flashcards on statistical tests, created by Tyde D on 11/12/2018.
- Created by
- Tyde D
- Views
- 0
- Rating