---
title: "Test and effect size details"
output:
  rmarkdown::html_vignette:
    toc: true
    toc_depth: 4
    keep_md: true
vignette: >
  %\VignetteIndexEntry{Test and effect size details}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r}
#| label = "setup",
#| message = FALSE,
#| warning = FALSE,
#| include = FALSE,
#| echo = FALSE
source("setup.R")
```

This vignette can be cited as:

```{r citation, echo=FALSE, comment = ""}
citation("statsExpressions")
```

## Introduction

This article describes the data that `{statsExpressions}` functions expect, and
summarizes, for each function, the statistical tests carried out, the effect
sizes returned, and the underlying functions used to compute them.

Abbreviations used: CI = Confidence Interval

## Data requirements

All functions expect data in **long (tidy) format** — one row per observation.
A few additional requirements are worth noting:

- **Within-subjects (repeated measures) designs**: The data must contain exactly
  *one* observation per subject per condition (a complete, balanced block design).
  If you have multiple trials per subject-condition cell, aggregate them first
  (e.g., by taking the mean) before passing the data.
  You can verify this with `table(data$subject, data$condition)` — every cell
  should equal `1`.

- **`subject.id` argument**: For within-subjects designs, always specify
  `subject.id` explicitly. If omitted, the function pairs observations by
  row order within each condition, so any data that is not already sorted
  identically within every condition level can produce silently incorrect
  paired tests — even with exactly two conditions and no missing values.

- **Missing data**: Missing values are removed internally. For within-subjects
  designs, any subject who has `NA` in *any* condition is removed entirely,
  ensuring a balanced design is maintained. For between-subjects designs, only
  the rows with `NA` are removed, unless `subject.id` is also supplied, in
  which case all rows of a subject with an `NA` are removed.

## Summary of functionality

```{r child="../man/rmd-fragments/functionality.Rmd"}
```

## Summary of tests and effect sizes

```{r child="../man/rmd-fragments/summary_intro.Rmd"}
```

### `centrality_description()`

```{r child="../man/rmd-fragments/centrality_description.Rmd"}
```

### `oneway_anova()`

```{r child="../man/rmd-fragments/oneway_anova.Rmd"}
```

### `two_sample_test()`

```{r child="../man/rmd-fragments/two_sample_test.Rmd"}
```

### `one_sample_test()`

```{r child="../man/rmd-fragments/one_sample_test.Rmd"}
```

### `corr_test()`

```{r child="../man/rmd-fragments/corr_test.Rmd"}
```

### `contingency_table()`

```{r child="../man/rmd-fragments/contingency_table.Rmd"}
```

### `pairwise_comparisons()`

```{r child="../man/rmd-fragments/pairwise_comparisons.Rmd"}
```

### `pairwise_contingency_table()`

```{r child="../man/rmd-fragments/pairwise_contingency_table.Rmd"}
```

### `meta_analysis()`

```{r child="../man/rmd-fragments/meta_analysis.Rmd"}
```

## Effect size interpretation

See `{effectsize}`'s interpretation functions to check different rules/conventions
to interpret effect sizes:

<https://easystats.github.io/effectsize/reference/index.html#section-interpretation>

## References

  - For parametric and non-parametric effect sizes:
    <https://easystats.github.io/effectsize/articles/>

  - For robust effect sizes:
    <https://CRAN.R-project.org/package=WRS2/vignettes/WRS2.pdf>

  - For Bayesian posterior estimates:
    <https://easystats.github.io/bayestestR/articles/bayes_factors.html>

## Suggestions

If you find any bugs or have any suggestions/remarks, please file an issue on GitHub:
<https://github.com/IndrajeetPatil/statsExpressions/issues>