Questions and lessons can't be printed or saved as a PDF.
Given the number of groups, pairing and normality, which test compares this numeric outcome?
More, if you want it
- Tell from repeated by the number of groups compared at once.
- Tell from by whether the same people appear at every time point.
- Tell from the by whether the same people are ranked repeatedly.
What does each patient in the study give?Tap a term: its definition, what you can do with it, and how often the exam asks it.
Definition: a test that compares means of a normal score between two independent groups.
You can: Tell the unpaired t-test from the paired t-test by whether the two sets of scores come from different people.
Definition: the rank-based twin of the unpaired t-test, for skewed scores in two groups.
Definition: a test that compares means in one group measured twice, with normal data.
You can: Tell the unpaired t-test from the paired t-test by whether the two sets of scores come from different people.
Definition: the rank-based twin of the paired t-test, for skewed paired data.
Definition: a test that compares means of a normal outcome across three or more independent groups.
You can: Tell one-way ANOVA from repeated t-tests by the number of groups compared at once.
Definition: the rank-based twin of one-way ANOVA, for ordinal or skewed data in three or more groups.
You can: Tell Friedman's test from the Kruskal-Wallis test by whether the same people are ranked repeatedly.
Definition: a test that compares means of one measure taken at three or more time points.
You can: Tell repeated measures ANOVA from one-way ANOVA by whether the same people appear at every time point.
Definition: the rank-based twin of repeated measures ANOVA, for ranks from the same people.
You can: Tell Friedman's test from the Kruskal-Wallis test by whether the same people are ranked repeatedly.
Definition: a test that compares counts of people in categories laid out as a table.
You can: Tell the chi-squared test from a t-test by whether each patient gives a category or a score.
Each card is one term. The grey line is what the patient says, which is what the question stem gives.
Summary
Choose a statistical test by asking about outcome type, number of groups, pairing, and whether the data are normal.
Choose a statistical test by asking about outcome type, number of groups, pairing, and whether the data are normal. An outcome is the result recorded for each person. The compares counts of people in categories.
Categorical data record presence or absence of an attribute. Continuous outcomes give a quantitative score along a continuum of intensity or frequency. Rating scales with ordered points, such as a Likert scale for symptom severity, give ordinal data. Normally distributed traits suit parametric tests. Skewed or ordinal data push the choice towards non-parametric tests, which work on ranks. A categorical outcome is a category, such as recovered or unwell. A continuous outcome is a score along a scale. An ordinal outcome is an ordered rating, such as a Likert scale. Normal data form a symmetrical bell shape. Skewed data lean to one side.
Harriet Lockwood is a statistician in a psychiatry trials unit. Harriet reviews a depression trial. Harriet says, "Each patient either recovered or did not, so I only have counts."
A averages a score for each group. A count of recovered patients gives no score to average. Counts of people in categories go to the chi-squared test.
6 more sections on Choosing a statistical test: t-tests, ANOVA, Mann-Whitney, Wilcoxon, Kruskal-Wallis and more
Log in free to read the rest: the full mental model, exam recognition, and the look-alikes that trip people up. Your daily session is free, always.
mrcpsych.co.uk · An independent study tool · not affiliated with the Royal College of Psychiatrists