---
title: "Getting started with aridagri"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Getting started with aridagri}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r setup, include = FALSE}
knitr::opts_chunk$set(collapse = TRUE, comment = "#>", fig.width = 6.5,
                      fig.height = 4, dpi = 96)
has_gg <- requireNamespace("ggplot2", quietly = TRUE)
library(aridagri)
```

aridagri analyses the field experiments that agronomists and breeders run
every season. Every function takes a data frame and column names, prints a
report, and returns an object you can reuse. Set `verbose = FALSE` to keep
only the object.

The examples use six simulated datasets that ship with the package
(`mothbean_rbd`, `clusterbean_spd`, `pearlmillet_met`, `chickpea_augmented`,
`cowpea_alpha`, `mungbean_repeated`). They mimic arid-zone trials but are not
real experimental results.

## 1. One trait in a randomized block design

```{r rbd}
res <- anova_rbd(mothbean_rbd, response = "seed_yield",
                 treatment = "treatment", block = "block",
                 posthoc = "duncan")
```

The report gives the ANOVA, SEm, SEd, CD at your `alpha`, CV, the relative
efficiency of blocking, and treatment means with letters. Missing plots are
handled by least squares: treatments are tested after blocks and every pair
of means is compared with its own standard error.

```{r rbd-plot, eval = has_gg}
plot(res, ylab = "Seed yield (kg/ha)")
```

## 2. The table for your paper: many traits at once

```{r traits}
tab <- anova_traits(mothbean_rbd,
                    traits = c("plant_height", "pods_per_plant",
                               "test_weight", "seed_yield", "harvest_index"),
                    design = "rbd", treatment = "treatment", block = "block",
                    posthoc = "duncan")
```

Letters are dropped and CD is shown as NS when the F-test is not
significant. `export_results(tab, "table.xlsx")` writes the table and the
underlying statistics to Excel.

## 3. Split-plot designs with all four CDs

```{r spd}
sp <- anova_spd(clusterbean_spd, response = "seed_yield",
                main_plot = "irrigation", sub_plot = "nitrogen",
                replication = "rep")
```

The SE table covers the four comparisons of Gomez and Gomez (1984),
including main-plot means at the same sub-plot level, which mixes Error(a)
and Error(b) and uses the weighted t value. The same engine runs the
split-plot designs with factorial main or sub-plots, the split-split-plot and
the combined analyses over years.

```{r spd-plot, eval = has_gg}
plot_interaction(sp, ylab = "Seed yield (kg/ha)")
```

## 4. Several environments: pooled analysis, stability and GGE biplot

```{r pooled}
pool <- anova_rbd_pooled(pearlmillet_met, response = "grain_yield",
                         treatment = "genotype", environment = "environment",
                         block = "rep", verbose = FALSE)
pool$bartlett_test[c("chi_sq", "p_value")]
pool$pooled_anova
```

```{r stability}
st <- stability_analysis(pearlmillet_met, genotype = "genotype",
                         environment = "environment", replication = "rep",
                         trait = "grain_yield", verbose = FALSE)
st$eberhart[, c("Genotype", "Mean", "bi", "p_bi", "S2di", "p_S2di")]
st$ammi$anova
```

```{r stability-plot, eval = has_gg}
plot_stability(st, "eberhart")
plot_stability(st, "ammi2")
```

```{r gge, eval = has_gg}
gg <- gge_biplot(pearlmillet_met, "genotype", "environment", "grain_yield",
                 verbose = FALSE)
gg$which_won_where
plot(gg, type = "which_won_where")
```

## 5. Germplasm evaluation: augmented and alpha lattice designs

```{r augmented}
au <- anova_augmented(chickpea_augmented, response = "seed_yield",
                      genotype = "entry", block = "block",
                      check_names = c("C1", "C2", "C3", "C4"),
                      verbose = FALSE)
head(au$test_vs_checks[, c("Entry", "Adjusted_mean", "Pct_over_best_check",
                           "Sig_vs_best")])
au$genetic_parameters
```

```{r alpha}
al <- anova_alpha_lattice(cowpea_alpha, response = "grain_yield",
                          genotype = "genotype", replication = "rep",
                          block = "block", verbose = FALSE)
al$efficiency
head(al$genotype_means)
```

## 6. Repeated measurements on the same plots or animals

```{r repeated}
rm <- anova_repeated(mungbean_repeated, response = "spad",
                     treatment = "treatment", time = "das", block = "block",
                     verbose = FALSE)
rm$sphericity
rm$corrected
rm$covariance_models
```

```{r repeated-plot, eval = has_gg}
plot(rm, ylab = "SPAD reading")
```

## 7. Agronomic indices

```{r indices}
thermal_indices(tmax = c(31, 33, 35, 36, 38), tmin = c(16, 17, 19, 20, 22),
                base_temp = 10, sunshine_hours = rep(9.5, 5),
                day_length = day_length(28, 70:74), crop_yield = 1250)

nue_calculate(yield_fertilized = c(1450, 1720, 1850), yield_control = 1200,
              nutrient_applied = c(20, 40, 60),
              nutrient_uptake_fert = c(48, 61, 68),
              nutrient_uptake_ctrl = 35)

nr <- nutrient_response(dose = c(0, 20, 40, 60, 80, 100),
                        yield = c(1100, 1350, 1520, 1610, 1650, 1660),
                        model = "all", price_output = 60,
                        price_nutrient = 15)
```

```{r response-plot, eval = has_gg}
plot_response(nr, c(0, 20, 40, 60, 80, 100),
              c(1100, 1350, 1520, 1610, 1650, 1660))
```

## 8. Letters from a published table

When you only have treatment means, the error mean square and its degrees of
freedom, `compare_means()` gives the letters:

```{r letters}
compare_means(c(T1 = 1180, T2 = 1265, T3 = 1342, T4 = 1410, T5 = 1238),
              mse = 3600, df_error = 12, reps = 4, method = "duncan")
```

## References

Gomez, K.A. and Gomez, A.A. (1984). *Statistical Procedures for Agricultural
Research*, 2nd ed. Wiley, New York.