# Compare groups

From the `Analysis`

menu choose `Compare groups`

, and choose the dependent variable(s) you want to analyze, choose the `Grouping variable`

, then hit OK.

In the `Display options...`

dialog, the minimum and maximum of the y-axis, and (new in v2.4) the way groups are displayed can be set.

The following results will be calculated (see also the common elements of the results):

## Raw data

Result | For interval variables | For ordinal variables | For nominal variables |
---|---|---|---|

Sample size | Number of observed cases | Number of observed cases | Number of observed cases |

Graphs of the raw data | Plot with individual data | Plot with individual data (order of the values) | Mosaic plot |

## Sample properties

Result | For interval variables | For ordinal variables | For nominal variables |
---|---|---|---|

Descriptive data | Means, Standard deviations, Maximums, Upper quartiles, Medians, Lower quartiles, Minimums | Maximums, Upper quartiles, Medians, Lower quartiles, Minimums | Variation ratio, Contingency table (case count), Contingency table (percentage) |

Standardized effect size | For two groups Cohen’s d and eta-squared | Cramér’s V | |

Graphs of the data | Box plot with individual data | Box plot with individual data (order of the values) |

## Population properties

Result | For interval variables | For ordinal variables | For nominal variables |
---|---|---|---|

Population parameters numerically | Mean with 95% CI | Median and its confidence interval | Contingency table (confidence interval, multinomial proportions with Goodman method) |

Graphs of the population parameters | Graph with mean and 95% CI of the mean | ||

Standardized effect size | * For two groups, Hedges’ g and its confidence interval * If one-way ANOVA was run, then omega-squared effect size. |
||

Hypothesis test for two groups and sensitivity power analyses | When one of the groups includes a single case, if control group is normal (measured with Shapiro-Wilk test) then modified t-test, otherwise Mann-Whitney test. If the `Single case slope` is set, then slope comparison.For comparing two groups if the data are normal (measured with Shapiro-Wilk test) and homoscedastic (measured with Levene’s test), then two-sample t-test and (new in v2.3) Bayesian two-samples t-test For non-normal groups, Mann-Whitney test For normal but heteroscedastic groups, Welch’s t-test Confidence interval of the difference of the groups For two-sample t-test sensitivity power analyses |
Mann-Whitney test | Chi-squared test and sensitivity power analyses |

Hypothesis test for more than two groups and sensitivity power analyses | If the data are normal (measured with Shapiro-Wilk test) and homoscedastic (measured with Levene’s test) one-way ANOVA If ANOVA is significant, post hoc Tukey’s HSD tests Otherwise, Kruskal-Wallis test If the Kruskal-Wallis test is significant, post hoc Dunn’s test For one-way ANOVA sensitivity power analyses displayed as eta-square and as f. |
Kruskal-Wallis test If the Kruskal-Wallis test is significant, post hoc Dunn’s test |
Chi-squared test and sensitivity power analyses |

Hypothesis test for two or more grouping variables | Two-way ANOVA or multi-way ANOVA |