### Install ggeffects from r-universe

Source: https://github.com/strengejacke/ggeffects/blob/master/README.rmd

Installation command for the development version hosted on r-universe.

```R
install.packages("ggeffects", repos = "https://strengejacke.r-universe.dev")
```

--------------------------------

### Install ggeffects from GitHub

Source: https://github.com/strengejacke/ggeffects/blob/master/README.rmd

Installation command for the development version using the remotes package.

```R
remotes::install_github("strengejacke/ggeffects")
```

--------------------------------

### Install ggeffects from CRAN

Source: https://github.com/strengejacke/ggeffects/blob/master/README.rmd

Standard installation command for the stable release of the package.

```R
install.packages("ggeffects")
```

--------------------------------

### Install latest development version

Source: https://github.com/strengejacke/ggeffects/blob/master/README.rmd

Convenience function to install the latest development version from r-universe.

```R
ggeffects::install_latest()
```

--------------------------------

### Display Pairwise Comparisons from brglm

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/brglm.md

Use this to print pairwise comparisons generated by ggpredict for brglm models. Contrasts are presented as probabilities in %-points.

```R
print(comp)
```

--------------------------------

### Subset ggeffects objects

Source: https://github.com/strengejacke/ggeffects/blob/master/NEWS.md

Demonstrates how to subset ggeffects objects using standard data frame indexing syntax.

```R
gge <- ggpredict(model, "x1")
gge[c(1:2)]
```

--------------------------------

### Print Compact Table of Predicted Values

Source: https://github.com/strengejacke/ggeffects/blob/master/README.md

Use `print()` with `collapse_table = TRUE` and `collapse_ci = TRUE` for a more compact display of predicted values and their confidence intervals.

```r
print(result, collapse_table = TRUE, collapse_ci = TRUE)
```

--------------------------------

### Retrieve package citation

Source: https://github.com/strengejacke/ggeffects/blob/master/README.rmd

Displays citation information for the ggeffects package.

```r
citation('ggeffects')
```

--------------------------------

### Print pairwise comparison output

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/test_predictions_ggeffects.md

Displays the results of pairwise comparisons for model factors.

```R
print(out1)
```

```R
print(out)
```

--------------------------------

### Load Libraries and Data for Prediction

Source: https://github.com/strengejacke/ggeffects/blob/master/README.md

Loads the ggeffects, splines, and datawizard libraries, along with the efc dataset. Converts specified columns to factors for model compatibility.

```r
library(ggeffects)
library(splines)
library(datawizard)
data(efc, package = "ggeffects")
efc <- to_factor(efc, c("c161sex", "e42dep"))
```

--------------------------------

### Pairwise Comparisons with Grouping by Sex

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/test_predictions_emmeans.md

This snippet shows how to perform pairwise comparisons while grouping the results by a categorical variable, such as 'sex', using the emmeans engine. This is useful for examining differences within subgroups.

```R
print(test_predictions(m, c("time", "coffee"), by = "sex", engine = "emmeans"))
```

--------------------------------

### Calculate interaction contrasts with test_predictions

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/test_predictions_emmeans.md

Computes interaction contrasts for a model using the emmeans engine. Requires a fitted model object and specified terms.

```R
print(test_predictions(m, terms = c("x1", "x2", "x3"), engine = "emmeans",
      test = "interaction"))
```

--------------------------------

### Generating Representative Values

Source: https://context7.com/strengejacke/ggeffects/llms.txt

Use pretty_range or bracket-notation shortcuts in predict_response to define specific values for continuous predictors.

```r
library(ggeffects)
data(efc, package = "ggeffects")

# Pretty range for a numeric variable
pretty_range(efc$c12hour)
#> [1]   0  20  40  60  80 100 120 140 160 180

# Specify number of intervals
pretty_range(1:1000, n = 5)
pretty_range(1:1000, length = 10)

# Using special terms shortcuts in predict_response
fit <- lm(barthtot ~ c12hour + neg_c_7, data = efc)

# Mean +/- 1 SD
predict_response(fit, terms = "c12hour [meansd]")

# Quartiles
predict_response(fit, terms = "c12hour [quart2]")

# All unique values
predict_response(fit, terms = "c12hour [all]")

# Min/Max
predict_response(fit, terms = "c12hour [minmax]")

# Sample of values (useful for random effects)
predict_response(fit, terms = "c12hour [sample=10]")

# Custom function transformation
predict_response(fit, terms = "c12hour [exp]")
```

--------------------------------

### Print Model Predictions with Default CI Format

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print.md

Prints model predictions with confidence intervals enclosed in parentheses. This is the default behavior for the `print()` function when applied to ggeffects objects.

```r
print(pr, group_name = FALSE, collapse_ci = TRUE)
```

--------------------------------

### Print Predicted Values for Factors

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print.md

Use this to print predicted values for factors from a model. The output shows predicted values and their confidence intervals for different levels of a factor.

```r
print(out, group_name = FALSE)
```

--------------------------------

### Print glmmTMB and orderedbeta predictions

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/glmmTMB.md

Displays the predicted proportions and confidence intervals for the specified model objects.

```R
print(out1)
```

```R
print(out2)
```

--------------------------------

### Predicting from Zero-Inflated Models

Source: https://context7.com/strengejacke/ggeffects/llms.txt

Use the type argument to specify whether to predict the response or the zero-inflation probability in zero-inflated models.

```r
library(glmmTMB)
data(Salamanders, package = "glmmTMB")
fit_zi <- glmmTMB(count ~ spp + mined + (1 | site),
                   ziformula = ~mined,
                   family = poisson,
                   data = Salamanders)

# Predictions conditioning on zero-inflation
predict_response(fit_zi, terms = "spp", type = "zero_inflated")

# Predicted zero-inflation probability
predict_response(fit_zi, terms = "mined", type = "zi_prob")
```

--------------------------------

### Print Predicted Probabilities with ggpredict

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/clm2.md

Displays predicted probabilities from a ggpredict object. This output is useful for understanding how model predictions vary across different levels of predictor variables.

```r
print(p)
```

--------------------------------

### Print Subsets of Predicted Values

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print_subsets.md

Use subsetting with `c()` to select and print specific rows of predicted values and their confidence intervals. This is useful for inspecting particular data points.

```r
print(gge[c(1:2, 4:6)])
```

--------------------------------

### Pairwise Comparisons without Grouping

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/test_predictions_emmeans.md

Use this to generate pairwise comparisons for specified variables using the emmeans engine. Ensure the model 'm' is appropriately defined.

```R
print(test_predictions(m, c("time", "coffee"), engine = "emmeans"))
```

--------------------------------

### ggaverage with collapsed tables

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/polr.md

Print the ggaverage output for a polr model with collapsed tables for a more compact view. This option is useful for summarizing predictions across multiple categories and predictors in a single table.

```r
print(pr, collapse_tables = TRUE)
```

--------------------------------

### Perform Pairwise Comparisons with test_predictions

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/test_predictions-margin.md

Calculates pairwise comparisons for specified model terms. Contrasts are presented as probabilities in percentage points.

```R
print(test_predictions(m, c("parfam [green, lib]", "year [1980, 2020]")))
```

```R
print(test_predictions(m, c("parfam [green, lib]", "year [1980, 2020]"),
margin = "marginalmeans"))
```

```R
print(test_predictions(m, c("parfam [green, lib]", "year [1980]")))
```

```R
print(test_predictions(m, c("parfam [green, lib]", "year [1980]"), margin = "marginalmeans"))
```

```R
print(test_predictions(m, c("parfam [green, lib]", "year")))
```

```R
print(test_predictions(m, c("parfam [green, lib]", "year"), margin = "marginalmeans"))
```

--------------------------------

### Print Model Predictions with Custom CI Brackets

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print.md

Prints model predictions with confidence intervals enclosed in square brackets. This is achieved by specifying the `ci_brackets` argument in the `print()` function.

```r
print(pr, group_name = FALSE, collapse_ci = TRUE, ci_brackets = c("[", "]"))
```

--------------------------------

### Collapse Tables in ggpredict

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print.md

Collapses tables for cleaner output and limits the number of rows displayed.

```R
print(ggpredict(m, c("Petal.Length", "Species")), collapse_tables = TRUE, n = 3)
```

--------------------------------

### Print ggpredict output

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/list_terms.md

Displays the predicted values for the model terms.

```R
print(out)
```

--------------------------------

### Plot predictions using the built-in plot method

Source: https://github.com/strengejacke/ggeffects/blob/master/README.rmd

Generates a default visualization for the predicted response object.

```r
mydf <- predict_response(fit, terms = "c12hour")
plot(mydf)
```

--------------------------------

### Configuring value labels and precision in ggpredict

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print_digits.md

Toggle the display of factor labels and adjust numeric precision for categorical predictors.

```R
print(ggpredict(fit, "e42dep"), value_labels = FALSE)
```

```R
print(ggpredict(fit, "e42dep"), value_labels = TRUE)
```

```R
print(ggpredict(fit, "e42dep"), value_labels = TRUE, digits = 4)
```

--------------------------------

### Visualize Adjusted Predictions with ggplot2

Source: https://context7.com/strengejacke/ggeffects/llms.txt

Use the plot() method to create publication-ready ggplot2 figures from ggeffects objects. Supports customization of confidence bands, facets, color palettes, and raw data points. Can also show partial residuals and use log scales for the y-axis.

```r
library(ggeffects)
library(ggplot2)
data(efc, package = "ggeffects")

efc$c172code <- as.factor(efc$c172code)
fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)

# Basic plot
mydf <- predict_response(fit, terms = "c12hour")
plot(mydf)
```

```r
# Plot with raw data points
plot(mydf, show_data = TRUE)
```

```r
# Customize appearance
plot(mydf,
     colors = "metro",           # Use metro color palette
     show_ci = TRUE,             # Show confidence intervals
     ci_style = "ribbon",        # Ribbon style for CI (or "errorbar", "dash", "dot")
     line_size = 1.2,
     dot_size = 3)
```

```r
# Multi-group predictions
mydf2 <- predict_response(fit, terms = c("c12hour", "c172code"))
plot(mydf2)
```

```r
# Faceted plot
plot(mydf2, facets = TRUE)
```

```r
# Three-way interaction with automatic faceting
mydf3 <- predict_response(fit, terms = c("c12hour", "c172code", "c161sex"))
plot(mydf3)
```

```r
# Black and white plot
plot(mydf2, colors = "bw")
```

```r
# Greyscale plot
plot(mydf2, colors = "gs")
```

```r
# Add partial residuals for model diagnostics
plot(mydf, show_residuals = TRUE, show_residuals_line = TRUE)
```

```r
# Log scale for y-axis (useful for binomial models)
plot(mydf, log_y = TRUE)
```

```r
# Show available color palettes
show_palettes()
```

--------------------------------

### Print hypothesis test results

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print_test_predictions.md

Displays the hypothesis test object. Use this to view pairwise comparisons and statistical metrics.

```R
print(ht)
```

```R
print(ht, table_width = Inf)
```

```R
print(ht)
```

--------------------------------

### Print Hypothesis Test with Collapsed CI

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print_test_predictions.md

Prints hypothesis test results with collapsed confidence intervals. This format can simplify the presentation of pairwise comparison results.

```R
print(out)
```

```R
print(out, collapse_ci = TRUE)
```

--------------------------------

### Print ggpredict output

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/avg_predictions.md

Displays the average predicted values for a model with specific conditions.

```R
print(out1)
```

--------------------------------

### Print Predicted Values with Custom Formatting

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print.md

Displays predicted values with group names enabled and confidence intervals collapsed into a single column using custom brackets.

```R
print(pr, group_name = TRUE, collapse_ci = TRUE, ci_brackets = c("[", "]"))
```

--------------------------------

### Print Test Predictions with Conditioned Values

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/test_predictions-mixed.md

Use test_predictions to generate pairwise comparisons for specified terms in a mixed model. The output displays contrasts, confidence intervals, and p-values. Contrasts are presented as counts.

```r
print(test_predictions(fit, terms = c("e16sex", "c172code")))
```

--------------------------------

### Test Predictions for 3-way Interaction using Emmeans

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/test_predictions_emmeans.md

Use this function to perform pairwise comparisons for a 3-way interaction when the emmeans engine is specified. Ensure the model 'm' and terms are correctly defined.

```R
print(test_predictions(m, terms = c("x1", "x2", "x3"), engine = "emmeans"))
```

--------------------------------

### Print Predicted Values with ggpredict

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print.md

Displays predicted values for specified terms in a statistical model. By default, not all rows are shown. Use `n = Inf` to display all rows.

```R
print(ggpredict(fit, terms = c("c12hour", "neg_c_7", "c161sex")))
```

```R
print(ggpredict(fit, terms = c("c12hour", "neg_c_7", "c161sex")), n = Inf)
```

--------------------------------

### Generate Predictions with glmmTMB

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print_test_predictions.md

Generates predictions from a glmmTMB model for a specific factor ('gear'). The output includes pairwise comparisons, confidence intervals, and p-values, presented as proportions.

```R
print(test_predictions(m, "gear"))
```

--------------------------------

### Visualize predictions with ggplot2

Source: https://github.com/strengejacke/ggeffects/blob/master/README.rmd

Uses the output of predict_response to create a line plot with confidence ribbons.

```r
library(ggplot2)
mydf <- predict_response(fit, terms = "c12hour")
ggplot(mydf, aes(x, predicted)) +
  geom_line() +
  geom_ribbon(aes(ymin = conf.low, ymax = conf.high), alpha = 0.1)
```

--------------------------------

### Extracting Plot Labels and Data

Source: https://context7.com/strengejacke/ggeffects/llms.txt

Helper functions allow extraction of titles and labels from ggeffects objects for custom ggplot2 visualizations.

```r
library(ggeffects)
library(ggplot2)
data(efc, package = "ggeffects")

efc$c172code <- as.factor(efc$c172code)
fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)

mydf <- predict_response(fit, terms = c("c12hour", "c161sex", "c172code"))

# Extract titles and labels
get_title(mydf)           # Response variable name/label
get_x_title(mydf)         # X-axis title (first focal term)
get_y_title(mydf)         # Y-axis title (response)
get_legend_title(mydf)    # Legend title (grouping variable)
get_x_labels(mydf)        # X-axis labels (if categorical)

# Custom ggplot with extracted labels
ggplot(mydf, aes(x = x, y = predicted, colour = group)) +
  geom_line() +
  geom_ribbon(aes(ymin = conf.low, ymax = conf.high, fill = group), alpha = 0.1) +
  facet_wrap(~facet) +
  labs(
    x = get_x_title(mydf),
    y = get_y_title(mydf),
    colour = get_legend_title(mydf),
    title = get_title(mydf)
  )

# Convert to data frame with original variable names
as.data.frame(mydf, terms_to_colnames = TRUE)
```

--------------------------------

### Generate Plot with ggplot2

Source: https://github.com/strengejacke/ggeffects/blob/master/README.md

Create a line plot of predicted values using ggplot2, with facets for different categories. Requires the `result` object and `aes` mappings for x, predicted values, color, and facets.

```r
ggplot(result, aes(x = x, y = predicted, colour = group)) +
  geom_line() +
  facet_wrap(~facet)
```

--------------------------------

### Format ggaverage output

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/polr.md

Format the output of ggaverage for a polr model into a more readable table. This includes columns for the predictor ('Infl'), predicted probability, confidence interval, and response category groups.

```r
format(pr)
```

--------------------------------

### Compute Average Counterfactual Predictions

Source: https://context7.com/strengejacke/ggeffects/llms.txt

Use ggaverage to compute average counterfactual predictions by averaging over all observations. This is useful for generalization to populations. Supports GLM models for predictions on the response scale.

```r
library(ggeffects)
data(efc, package = "ggeffects")

fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)

# Average predictions (counterfactual)
ggaverage(fit, terms = "c12hour")
```

```r
# With specific values
ggaverage(fit, terms = "c12hour [0, 50, 100, 150]")
```

```r
# Using weighted predictions
ggaverage(fit, terms = "c172code", weights = "c12hour")
```

```r
# For GLM models - predictions on response scale
fit_glm <- glm(tot_sc_e ~ c12hour + neg_c_7 + c161sex, data = efc, family = poisson())
ggaverage(fit_glm, terms = "c12hour")
```

--------------------------------

### Test Pairwise Comparisons and Contrasts

Source: https://context7.com/strengejacke/ggeffects/llms.txt

Use test_predictions to test differences between adjusted predictions for statistical significance. Supports pairwise comparisons, contrasts, and marginal effects. P-values can be adjusted using methods like 'tukey' or 'bonferroni'.

```r
library(ggeffects)
data(efc, package = "ggeffects")

efc$c172code <- as.factor(efc$c172code)
efc$c161sex <- as.factor(efc$c161sex)
levels(efc$c161sex) <- c("male", "female")

fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)

# Pairwise comparisons for a factor
test_predictions(fit, "c172code")
```

```r
# Test predictions from a ggeffects object
pred <- predict_response(fit, "c172code")
test_predictions(pred)
```

```r
# Test for slope/trend of continuous predictor
test_predictions(fit, "c12hour")
```

```r
# Interaction contrasts
fit2 <- lm(barthtot ~ c12hour + c161sex * c172code + neg_c_7, data = efc)
test_predictions(fit2, c("c161sex", "c172code"), test = "pairwise")
```

```r
# Contrasts by groups
test_predictions(fit2, "c172code", by = "c161sex")
```

```r
# Different test types
test_predictions(fit2, c("c161sex", "c172code"), test = "interaction")
test_predictions(fit2, "c172code", test = "consecutive")
test_predictions(fit2, "c172code", test = "contrast")
```

```r
# P-value adjustment
test_predictions(fit, "c172code", p_adjust = "tukey")
test_predictions(fit, "c172code", p_adjust = "bonferroni")
```

--------------------------------

### ggpredict with polr model

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/polr.md

Use ggpredict to obtain predicted probabilities for each category of the response variable from a polr model. The output shows predicted probabilities for 'Sat' across different levels of 'Infl', adjusted for other variables.

```r
print(pr)
```

--------------------------------

### Fit Linear Model for Prediction

Source: https://github.com/strengejacke/ggeffects/blob/master/README.md

Fits a linear model using lm() with interaction and spline terms. This model will be used to generate adjusted predictions.

```r
fit <- lm(barthtot ~ c12hour + bs(neg_c_7) * c161sex + e42dep, data = efc)
```

--------------------------------

### Calculate Interaction Terms

Source: https://context7.com/strengejacke/ggeffects/llms.txt

Use ggemmeans to calculate interaction terms for specified variables in a model.

```r
ggemmeans(fit, terms = c("c172code", "c161sex"))
```

--------------------------------

### Print Weighted Average Predictions

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print.md

Calculates and prints average predicted values with collapsed tables for complex models.

```R
print(ggaverage(m, c("Infl", "Type", "Sat")), collapse_tables = TRUE)
```

--------------------------------

### ggpredict - Predictions with Non-focal Terms at Mean/Reference

Source: https://context7.com/strengejacke/ggeffects/llms.txt

`ggpredict()` computes predictions where non-focal predictors are held at their mean (numeric) or reference level (factors). This represents a "typical" observation approach and is called internally when `margin = "mean_reference"`.

```APIDOC
## ggpredict - Predictions with Non-focal Terms at Mean/Reference

### Description
Computes predictions where non-focal predictors are held at their mean (numeric) or reference level (factors). This represents a "typical" observation approach and is called internally when `margin = "mean_reference"`.

### Method
`ggpredict(model, terms, typical = c("numeric" = "mean", "factor" = "mode"), ...)`

### Parameters
#### Path Parameters
- **model** (model object) - Required - The statistical model object.
- **terms** (character or list) - Required - The focal terms for which to compute predictions. Can be a single term, a vector of terms, or a list for interactions.
- **typical** (character vector) - Optional - Specifies how non-focal predictors are handled. Defaults to "mean" for numeric and "mode" for factor variables.

### Request Example
```r
library(ggeffects)
data(efc, package = "ggeffects")
fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)

# Basic usage - predictions for c12hour
mydf <- ggpredict(fit, terms = "c12hour")
print(mydf)

# Predictions with grouping variable
ggpredict(fit, terms = c("c12hour", "c172code"))

# Specify typical values for covariates
ggpredict(fit, terms = "c12hour", typical = c(numeric = "mean", factor = "mode"))

# Range of values with step size
ggpredict(fit, terms = "c12hour [0:100 by=10]")

# Using named list format for terms
ggpredict(fit, terms = list(c12hour = seq(0, 100, 20), c172code = c(1, 3)))

# Three-way interaction
ggpredict(fit, terms = c("c12hour", "c161sex", "c172code"))
```

### Response
#### Success Response (200)
Returns a data frame with columns for the focal term(s), predicted values, and confidence intervals (e.g., `95% CI`).

#### Response Example
```
# Predicted values of barthtot

c12hour | Predicted |       95% CI
----------------------------------
      4 |     74.61 | 72.67, 76.55
     12 |     74.05 | 72.18, 75.91
     22 |     73.33 | 71.48, 75.19
```
```

--------------------------------

### Use vector variables in terms within functions

Source: https://github.com/strengejacke/ggeffects/blob/master/NEWS.md

Shows how to pass a vector variable defined inside a function to the terms argument of ggpredict.

```R
foo <- function(data) {
  fit <- lm(barthtot ~ c12hour + c172code, data = data)
  v <- c(20, 50, 70)
  ggpredict(fit, terms = "c12hour [v]")
}
foo(efc)
```

--------------------------------

### Generate Plot with Base R Plot Function

Source: https://github.com/strengejacke/ggeffects/blob/master/README.md

Use the base R `plot()` function to generate a plot from the `result` object. This provides a simpler plotting alternative to ggplot2.

```r
plot(result)
```

--------------------------------

### Adjust Confidence Interval Levels

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print.md

Displays predicted values with different confidence interval levels.

```R
print(out)
```

```R
print(out)
```

--------------------------------

### Plot Marginal Effects with ggplot2

Source: https://github.com/strengejacke/ggeffects/blob/master/paper/paper.md

Visualize the marginal effects calculated by ggpredict using ggplot2. This approach allows for customization of the plot. Requires the ggplot2 library.

```R
library(ggplot2)
mydf <- ggpredict(fit, terms = "c12hour")
ggplot(mydf, aes(x, predicted)) + 
  geom_line() + 
  geom_ribbon(aes(ymin = conf.low, ymax = conf.high), alpha = .1)
```

--------------------------------

### Handle Mixed Effects Models

Source: https://context7.com/strengejacke/ggeffects/llms.txt

ggeffects supports mixed-effects models from lme4, glmmTMB, and other packages. Provides options for population-level ('fixed') and unit-level ('random') predictions, prediction intervals, and empirical/counterfactual marginal predictions.

```r
library(ggeffects)
library(lme4)

data(sleepstudy, package = "lme4")
fit <- lmer(Reaction ~ Days + (Days | Subject), data = sleepstudy)

# Population-level predictions (default)
predict_response(fit, terms = "Days", type = "fixed")
```

```r
# Unit-level predictions (include random effects)
predict_response(fit, terms = c("Days", "Subject [sample=5]"), type = "random")
```

```r
# Prediction intervals (include residual variance)
predict_response(fit, terms = "Days", interval = "prediction")
```

```r
# Using empirical/counterfactual approach for marginal predictions
predict_response(fit, terms = "Days", margin = "empirical")
```

--------------------------------

### ggemmeans - Estimated Marginal Means

Source: https://context7.com/strengejacke/ggeffects/llms.txt

`ggemmeans()` uses the emmeans package to compute estimated marginal means, where non-focal categorical predictors are marginalized over their levels. Called internally when `margin = "marginalmeans"`.

```APIDOC
## ggemmeans - Estimated Marginal Means

### Description
Uses the emmeans package to compute estimated marginal means, where non-focal categorical predictors are marginalized over their levels. Called internally when `margin = "marginalmeans"`.

### Method
`ggemmeans(model, terms, weights = "proportional", ...)`

### Parameters
#### Path Parameters
- **model** (model object) - Required - The statistical model object.
- **terms** (character or list) - Required - The focal terms for which to compute estimated marginal means.
- **weights** (character) - Optional - Specifies the weighting scheme for marginalization. Options include "proportional" (default), "equal", or a numeric vector.

### Request Example
```r
library(ggeffects)
data(efc, package = "ggeffects")

efc$c172code <- as.factor(efc$c172code)
fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)

# Estimated marginal means
ggemmeans(fit, terms = "c172code")

# With different weighting options
ggemmeans(fit, terms = "c172code", weights = "proportional")
ggemmeans(fit, terms = "c172code", weights = "equal")
```

### Response
#### Success Response (200)
Returns a data frame with columns for the focal term(s), estimated marginal means, and confidence intervals (e.g., `95% CI`).

#### Response Example
```
# Predicted values of barthtot

c172code | Predicted |       95% CI
-----------------------------------
       1 |     69.07 | 66.68, 71.46
       2 |     70.51 | 68.95, 72.07
       3 |     67.88 | 65.17, 70.59

Adjusted for:
* c12hour = 42.10
* neg_c_7 = 11.83
* c161sex =  1.00
```
```

--------------------------------

### predict_response - Main Function for Computing Adjusted Predictions

Source: https://context7.com/strengejacke/ggeffects/llms.txt

The primary function `predict_response()` computes marginal means and adjusted predictions at the margin of specific values or levels from focal model terms. It acts as a unified wrapper around specialized functions (`ggpredict()`, `ggemmeans()`, `ggaverage()`) with the `margin` argument controlling how non-focal predictors are marginalized.

```APIDOC
## predict_response - Main Function for Computing Adjusted Predictions

### Description
Computes marginal means and adjusted predictions at the margin of specific values or levels from focal model terms. It acts as a unified wrapper around specialized functions (`ggpredict()`, `ggemmeans()`, `ggaverage()`) with the `margin` argument controlling how non-focal predictors are marginalized.

### Method
`predict_response(model, terms, margin = "mean_reference", ...)`

### Parameters
#### Path Parameters
- **model** (model object) - Required - The statistical model object.
- **terms** (character or list) - Required - The focal terms for which to compute predictions. Can be a single term, a vector of terms, or a list for interactions.
- **margin** (character) - Optional - Specifies how non-focal predictors are marginalized. Options include "mean_reference" (default), "marginalmeans", "empirical", "weighted.mean".

### Request Example
```r
library(ggeffects)
data(efc, package = "ggeffects")
fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)

# Basic prediction for a single term
predict_response(fit, terms = "c12hour")

# Predictions with multiple focal terms (interactions)
predict_response(fit, terms = c("c12hour", "c172code"))

# Specify meaningful values using brackets
predict_response(fit, terms = "c12hour [30:60]")

# Use specific values for focal terms
predict_response(fit, terms = c("c12hour", "c172code [1,3]"))

# Hold non-focal terms at specific values with condition
predict_response(fit, terms = "c12hour [40:60]", condition = c(neg_c_7 = 20))

# Different marginalization methods
predict_response(fit, terms = "c12hour", margin = "mean_reference")
predict_response(fit, terms = "c12hour", margin = "marginalmeans")
predict_response(fit, terms = "c12hour", margin = "empirical")
```

### Response
#### Success Response (200)
Returns a data frame with columns for the focal term(s), predicted values, and confidence intervals (e.g., `95% CI`).

#### Response Example
```
# Predicted values of barthtot

c12hour | Predicted |       95% CI
----------------------------------
      4 |     74.61 | 72.67, 76.55
     12 |     74.05 | 72.18, 75.91
     22 |     73.33 | 71.48, 75.19
     36 |     72.33 | 70.43, 74.22
     49 |     71.40 | 69.43, 73.37
     70 |     69.90 | 67.78, 72.03
    100 |     67.76 | 65.35, 70.16
    168 |     62.91 | 59.57, 66.24

Adjusted for:
* neg_c_7  = 11.83
* c161sex  =     1
* c172code =     2
```
```

--------------------------------

### Predictions with Non-focal Terms at Mean/Reference using ggpredict

Source: https://context7.com/strengejacke/ggeffects/llms.txt

ggpredict() computes predictions by holding non-focal predictors at their mean (numeric) or reference level (factors). This function is called internally by predict_response() when margin = "mean_reference".

```R
library(ggeffects)
data(efc, package = "ggeffects")

fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)

# Basic usage - predictions for c12hour
mydf <- ggpredict(fit, terms = "c12hour")
mydf
```

```R
# Predictions with grouping variable
ggpredict(fit, terms = c("c12hour", "c172code"))
```

```R
# Specify typical values for covariates
ggpredict(fit, terms = "c12hour", typical = c(numeric = "mean", factor = "mode"))
```

```R
# Range of values with step size
ggpredict(fit, terms = "c12hour [0:100 by=10]")
```

```R
# Using named list format for terms
ggpredict(fit, terms = list(c12hour = seq(0, 100, 20), c172code = c(1, 3)))
```

```R
# Three-way interaction
ggpredict(fit, terms = c("c12hour", "c161sex", "c172code"))
```

--------------------------------

### Calculate and visualize predictions for multiple focal predictors

Source: https://github.com/strengejacke/ggeffects/blob/master/README.rmd

Handles up to four terms, allowing for grouping and faceting in the resulting plots.

```r
result <- predict_response(fit, terms = c("neg_c_7", "c161sex", "e42dep"))
# we want a more compact table, thus we use `print()` explicitly
print(result, collapse_table = TRUE, collapse_ci = TRUE)

ggplot(result, aes(x = x, y = predicted, colour = group)) +
  geom_line() +
  facet_wrap(~facet)
```

```r
plot(result)
```

--------------------------------

### Calculate Marginal Effects with ggpredict

Source: https://github.com/strengejacke/ggeffects/blob/master/paper/paper.md

Use ggpredict to calculate predicted values for a linear model. Requires the ggeffects and sjlabelled libraries. The 'terms' argument specifies the variable for which to calculate effects.

```R
library(ggeffects)
library(sjlabelled)
data(efc)
efc$c172code <- as_factor(efc$c172code)
fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)

ggpredict(fit, terms = "c12hour")
```

--------------------------------

### Plot Marginal Effects with plot() Method

Source: https://github.com/strengejacke/ggeffects/blob/master/paper/paper.md

Use the built-in plot() method for ggpredict objects to quickly generate a ggplot object representing the marginal effects. This method handles common plot characteristics.

```R
p <- ggpredict(fit, terms = c("c172code", "c161sex"))
plot(p)
```

--------------------------------

### Test pairwise comparisons for glmmTMB

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/windows/print_zero_inflation.md

Performs pairwise comparisons on the predicted counts from a zero-inflated model fitted with glmmTMB. Contrasts are presented as counts.

```r
print(test_predictions(out))
```

--------------------------------

### Applying Robust Standard Errors

Source: https://context7.com/strengejacke/ggeffects/llms.txt

Specify the vcov argument to use robust standard errors or custom variance-covariance matrices for inference.

```r
library(ggeffects)
data(efc, package = "ggeffects")

fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)

# Heteroscedasticity-consistent standard errors
predict_response(fit, terms = "c12hour", vcov = "HC3")

# Various HC types
predict_response(fit, terms = "c12hour", vcov = "HC0")
predict_response(fit, terms = "c12hour", vcov = "HC1")
predict_response(fit, terms = "c12hour", vcov = "HC4")

# Cluster-robust standard errors (requires clubSandwich)
# predict_response(fit, terms = "c12hour", vcov = "CR2", vcov_args = list(cluster = efc$cluster_var))

# Bootstrap standard errors
predict_response(fit, terms = "c12hour", vcov = "BS")

# Custom variance-covariance matrix
my_vcov <- sandwich::vcovHC(fit, type = "HC3")
predict_response(fit, terms = "c12hour", vcov = my_vcov)
```

--------------------------------

### Compute Adjusted Predictions with predict_response

Source: https://context7.com/strengejacke/ggeffects/llms.txt

Use predict_response() to calculate marginal means and adjusted predictions. It supports various marginalization methods via the 'margin' argument and allows specifying focal terms and conditions.

```R
library(ggeffects)
data(efc, package = "ggeffects")

# Fit a linear model
fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)

# Basic prediction for a single term
predict_response(fit, terms = "c12hour")
```

```R
# Predictions with multiple focal terms (interactions)
predict_response(fit, terms = c("c12hour", "c172code"))
```

```R
# Specify meaningful values using brackets
predict_response(fit, terms = "c12hour [30:60]")
```

```R
# Use specific values for focal terms
predict_response(fit, terms = c("c12hour", "c172code [1,3]"))
```

```R
# Hold non-focal terms at specific values with condition
predict_response(fit, terms = "c12hour [40:60]", condition = c(neg_c_7 = 20))
```

```R
# Different marginalization methods
predict_response(fit, terms = "c12hour", margin = "mean_reference")  # default
predict_response(fit, terms = "c12hour", margin = "marginalmeans")   # weighted average
predict_response(fit, terms = "c12hour", margin = "empirical")       # counterfactual
```

--------------------------------

### Pooling Predictions from Imputed Data

Source: https://context7.com/strengejacke/ggeffects/llms.txt

Use pool_predictions to combine results from multiple imputed datasets generated by the mice package.

```r
library(ggeffects)
library(mice)

# Create multiply imputed datasets
data("nhanes2", package = "mice")
imp <- mice(nhanes2, m = 5, printFlag = FALSE)

# Generate predictions for each imputed dataset
predictions <- lapply(1:5, function(i) {
  m <- lm(bmi ~ age + hyp + chl, data = complete(imp, action = i))
  predict_response(m, "age")
})

# Pool predictions
pooled <- pool_predictions(predictions)
pooled
#> # Predicted values of bmi
#>
#> age   | Predicted |       95% CI
#> --------------------------------
#> 20-39 |     27.37 | 24.89, 29.85
#> 40-59 |     28.64 | 26.03, 31.26
#> 60-99 |     26.99 | 23.43, 30.54

# Plot pooled predictions
plot(pooled)
```

--------------------------------

### Adjusting digits and row display in ggpredict

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/print_digits.md

Control the precision of numeric output and the number of rows displayed in the console.

```R
print(ggpredict(m, "Petal.Length"), digits = 5)
```

```R
print(ggpredict(m, "Petal.Length"), digits = 4, n = 3)
```

--------------------------------

### Predicted Probabilities with Covariate Adjustment

Source: https://github.com/strengejacke/ggeffects/blob/master/tests/testthat/_snaps/clm.md

This snippet generates predicted probabilities for an ordinal model, showing the effect of a covariate ('contact') across different levels of the outcome and predictor ('temp').

```r
print(p2)
```

--------------------------------

### Compute Estimated Marginal Means with ggemmeans

Source: https://context7.com/strengejacke/ggeffects/llms.txt

ggemmeans() calculates estimated marginal means using the emmeans package. Non-focal categorical predictors are marginalized over their levels. This function is called internally by predict_response() when margin = "marginalmeans".

```R
library(ggeffects)
data(efc, package = "ggeffects")

efc$c172code <- as.factor(efc$c172code)
fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)

# Estimated marginal means
ggemmeans(fit, terms = "c172code")
```

```R
# With different weighting options
ggemmeans(fit, terms = "c172code", weights = "proportional")
ggemmeans(fit, terms = "c172code", weights = "equal")
```

--------------------------------

### Calculate Adjusted Predictions for One Predictor

Source: https://github.com/strengejacke/ggeffects/blob/master/README.md

Calculates adjusted predictions for the 'barthtot' response variable based on the 'c12hour' predictor. Predictions are adjusted for other model variables.

```r
predict_response(fit, terms = "c12hour")
```