dotwhisker

https://fsolt.org/dotwhisker/
Admin
dotwhisker is an R package for quickly and easily generating dot-and-whisker plots of regression...
Tokens:27,912
Snippets:188
Trust Score:8.5
Update:4 months ago
Show doc for...
Tokens:
Raw
### Install Development Version of dotwhisker

Source: https://fsolt.org/dotwhisker/index.html

Install the latest development version of the dotwhisker package from GitHub. Ensure the 'remotes' package is installed first.

```r
if (!require("remotes")) install.packages("remotes"); remotes::install_github("fsolt/dotwhisker")
```

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

### Initialize Empty List for Estimates

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

Initializes an empty list to store regression estimates. This is a common starting point before populating it with data.

```R
Y1 <- vector(length=0,mode="list")
#estimates with law change (in the 4th to 6th columns)
Y1$estimate <- coef.matrix[,4:6]
##95% confidence intervals
Y1$lo <- coef.matrix[,4:6]-qnorm(0.975)*se.matrix[,4:6]
Y1$hi <- coef.matrix[,4:6]+qnorm(0.975)*se.matrix[,4:6]
##90% confidence intervals
Y1$lo1 <- coef.matrix[,4:6]-qnorm(0.95)*se.matrix[,4:6]
Y1$hi1 <- coef.matrix[,4:6]+qnorm(0.95)*se.matrix[,4:6]
##name the rows of Y1 estimate
rownames(Y1$estimate) <- varnames[-7] ##no intercept
```

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

### Install Latest Release Version of dotwhisker

Source: https://fsolt.org/dotwhisker/index.html

Use this command to install the latest stable release of the dotwhisker package from CRAN.

```r
install.packages("dotwhisker")
```

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

### Load Required Packages and Run Regression

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

Loads the dotwhisker and dplyr packages, then runs a linear regression model using the mtcars dataset. This setup is necessary for generating dot-and-whisker plots.

```r
#Package preload
library(dotwhisker)
library(dplyr)

# run a regression compatible with tidy
m1 <- lm(mpg ~ wt + cyl + disp + gear, data = mtcars)

```

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

### Example Usage of add_brackets with dwplot

Source: https://fsolt.org/dotwhisker/reference/add_brackets.html

Demonstrates how to use add_brackets to group predictors in a dot-and-whisker plot. The example first creates a linear model, defines two sets of brackets, generates a dwplot with custom relabeling and theming, and then applies the add_brackets function to the plot.

```R
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
m1 <- lm(mpg ~ wt + cyl + disp, data = mtcars)
two_brackets <- list(c("Engine", "Cylinder", "Displacement"),
                     c("Not Engine", "Intercept", "Weight"))

{dwplot(m1, show_intercept = TRUE) %>%
       relabel_predictors("(Intercept)" = "Intercept",
                          wt = "Weight",
                          cyl = "Cylinder",
                          disp = "Displacement") + 
       theme_bw() + xlab("Coefficient") + ylab("") +
       theme(legend.position="none") +
       geom_vline(xintercept = 0, colour = "grey50", linetype = 2)} %>%
   add_brackets(two_brackets)
```

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

### Rescale Regression Results with by_2sd

Source: https://fsolt.org/dotwhisker/reference/by_2sd.html

Example of using by_2sd to rescale regression results from a linear model for plotting. Requires the broom and dplyr libraries.

```R
library(broom)
library(dplyr)

data(mtcars)
m1 <- lm(mpg ~ wt + cyl + disp, data = mtcars)
m1_df <- tidy(m1) %>% by_2sd(mtcars) # create data frame of rescaled regression results

```

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

### Advanced Plotting with Multiple Models and Customization

Source: https://fsolt.org/dotwhisker/reference/dwplot.html

Generate a complex dot-and-whisker plot from multiple models created on the fly. This example includes relabeling predictors, applying a theme, customizing axes and title, and adjusting legend appearance.

```R
# Plot regression coefficients from multiple models on the fly
mtcars %>%
    split(.$am) %>%
    purrr::map(~ lm(mpg ~ wt + cyl + disp, data = .x)) %>%
    dwplot() %>%
    relabel_predictors(c(wt = "Weight", cyl = "Cylinders", disp = "Displacement")) +
    theme_bw() + xlab("Coefficient") + ylab("") +
    geom_vline(xintercept = 0, colour = "grey60", linetype = 2) +
    ggtitle("Predicting Gas Mileage, OLS Estimates") +
    theme(plot.title = element_text(face = "bold"),
          legend.position = c(.995, .99),
          legend.justification = c(1, 1),
          legend.background = element_rect(colour="grey80"),
          legend.title.align = .5) +
    scale_colour_grey(start = .4, end = .8,
                      name = "Transmission",
                      breaks = c("Model 0", "Model 1"),
                      labels = c("Automatic", "Manual"))
#> Warning: The `legend.title.align` argument of `theme()` is deprecated as of ggplot2
#> 3.5.0.
#> ℹ Please use theme(legend.title = element_text(hjust)) instead.
#> Warning: A numeric `legend.position` argument in `theme()` was deprecated in ggplot2
#> 3.5.0.
#> ℹ Please use the `legend.position.inside` argument of `theme()` instead.
```

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

### Integrating Model Fit Statistics with Relabeling and Brackets

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

This example demonstrates a workaround to combine model fit statistics with relabeled predictors and added brackets on a dot-whisker plot. It involves separately generating the plot with brackets and then appending the model fit statistics using `gridExtra` and `patchwork`.

```r
library(gridExtra)
library(patchwork)

three_brackets <- list(
    c("Overall", "Weight", "Weight"),
    c("Engine", "Cylinders", "Horsepower"),
    c("Transmission", "Gears", "Manual")
)

plot_brackets <- { dwplot(m3, 
           vline = geom_vline(
               xintercept = 0,
               colour = "grey60",
               linetype = 2
           )) |> # plot line at zero _behind_ coefs
        relabel_predictors(
            c(
                wt = "Weight",
                # relabel predictors
                cyl = "Cylinders",
                disp = "Displacement",
                hp = "Horsepower",
                gear = "Gears",
                am = "Manual"
            )
        ) + xlab("Coefficient Estimate") + ylab("") +
        ggtitle("Predicting Gas Mileage") 
} |> 
    add_brackets(three_brackets, fontSize = 0.3)

plot_brackets / tableGrob(
  dotwhisker:::dw_stats(
    m3,
    stats_digits = 2,
    stats_compare = FALSE
  ),
  rows = NULL,
  theme = ttheme_default(base_size = 3)
) +
  plot_layout(heights = c(5, -0.5, 1)) # the negative value is used to adjust the space between the plot and the model fits
```

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

### Generate Secret Weapon Plot from Tidy Data Frame

Source: https://fsolt.org/dotwhisker/reference/secret_weapon.html

This example demonstrates how to generate a 'secret weapon' plot from a tidy data frame of regression results. It first estimates linear models across different groups and then uses `secret_weapon` to visualize the 'carat' coefficient.

```R
library(dplyr)
library(broom)

# Estimate models across many samples, put results in a tidy data frame
by_clarity <- diamonds %>% group_by(clarity) %>%
 do(tidy(lm(price ~ carat + cut + color, data = .))) %>%
ungroup %>% rename(model = clarity)

# Generate a 'secret weapon' plot of the results of diamond size
secret_weapon(by_clarity, "carat")
```

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

### Advanced Customization: Ordering, Relabeling, and Theming

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

Customizes the plot by ordering models and variables, relabeling predictors, and applying custom ggplot2 themes and layers. This example also adds a vertical line at zero.

```r
dwplot(list(m1, m2, m3),
       vline = geom_vline(
           xintercept = 0,
           colour = "grey60",
           linetype = 2
       ),
       vars_order = c("am", "cyl", "disp", "gear", "hp", "wt"),
       model_order = c("Model 2", "Model 1", "Model 3")
       )  |>  # plot line at zero _behind_coefs
    relabel_predictors(
        c(
            am = "Manual",
            cyl = "Cylinders",
            disp = "Displacement",
            wt = "Weight",
            gear = "Gears",
            hp = "Horsepower"
        )
    ) +
    theme_bw(base_size = 4) + 
    # Setting `base_size` for fit the theme
    # No need to set `base_size` in most usage
    xlab("Coefficient Estimate") + ylab("") +
    geom_vline(xintercept = 0,
               colour = "grey60",
               linetype = 2) + 
    ggtitle("Predicting Gas Mileage") + 
    theme(
        plot.title = element_text(face = "bold"),
        legend.position = c(0.007, 0.01),
        legend.justification = c(0, 0),
        legend.background = element_rect(colour = "grey80"),
        legend.title = element_blank()
    ) 
```

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

### Building Tidy Data Frame Directly

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

Demonstrates how to build a tidy data frame of regression results directly, without first creating model objects, and relabeling predictors.

```r
# Run model on subsets of data, save results as tidy df, make a model variable, and relabel predictors
by_trans <- mtcars |
 group_by(am) |> # group data by trans
 do(broom::tidy(lm(mpg ~ wt + cyl + disp + gear, data = .))) |> # run model on each grp
 rename(model = am) |> # make model variable
 relabel_predictors(c(
 wt = "Weight",
 # relabel predictors
 cyl = "Cylinders",
 disp = "Displacement",
 gear = "Gear"
 ))

by_trans
```

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

### Initialize List for Estimates Without Law Change

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

Initializes and populates a list for regression estimates that do not include a law change, calculating confidence intervals.

```R
#estimates without law change
Y2 <- vector(length=0,mode="list")
Y2$estimate <- coef.matrix[,1:3]
Y2$lo <- coef.matrix[,1:3]-qnorm(.975)*se.matrix[,1:3]
Y2$hi <- coef.matrix[,1:3]+qnorm(.975)*se.matrix[,1:3]
Y2$lo1 <- coef.matrix[,1:3]-qnorm(.95)*se.matrix[,1:3]
Y2$hi1 <- coef.matrix[,1:3]+qnorm(.95)*se.matrix[,1:3]
rownames(Y2$estimate) <- varnames[-7]
```

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

### Prepare Regression Data Matrices

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

Initializes matrices for point estimates and standard errors, along with R-squared values and variable names, for multiple regression models. This is a precursor to plotting.

```R
library(grid)

##point estimates, in a n.variables, n.variables x n.models
coef.matrix <- matrix(c(-.039, NA, .048, -.133, .071, -.795, 1.47, 
                        -.036, NA, .036, -.142, .07, -.834, 1.70, 
                        -.051, NA, .017, .05, .011, -.532, .775, 
                        -.037, -.02, .047, -.131,.072, -.783, 1.45,
                        -.034, -.018, -.035, -.139, .071, -.822, 1.68,
                        -.05, -.023, .016,-.049, .013, -.521, .819),nr=7)
## R2 of the models
R2<-  c(0.910,  0.910,  0.940,  0.910,  0.910,  0.940)

##standard error matrix, n.variables x n.models
se.matrix <- matrix(c(.003, NA, .011, .013, .028, .056, .152, .003, NA, .012, .014, .029, .059, .171, .003, NA,
                      .01, .013, .024, .044, .124, .003, .005, .011, .013, .028, .055, .152, .003, .005, .021, .014,
                      .029, .059, .17, .003,.006, .01, .013, .024, .044, .127),nr=7)

##variable names
varnames<- c("% of county\nregistration", "Law change", "Log population", "Log median\nfamily income",
             "% population with\nh.s. education" ,"% population\nAfrican American" ,"Constant")


##exclude intercept
coef.matrix<-coef.matrix[-(7,)]
se.matrix<-se.matrix[-7,]

## each panel has at most six models, plotted in pairs.
## in each pair, solid circles will be the models with "law change" in the specification

```

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

### Basic Secret Weapon Plot

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

A basic example of a 'secret weapon' plot, which visualizes regression coefficients with confidence intervals. This is useful for presenting results from a single model.

```R
secret_weapon(by.clarity, var = "carat") +
    xlab("Estimated Coefficient (Dollars)") +
    ylab("Diamond Clarity") +
    ggtitle("Estimates for Diamond Size Across Clarity Grades") +
    theme(plot.title = element_text(face = "bold"))
```

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

### Create and Populate a Graphical Frame

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

This snippet demonstrates how to create a graphical frame (frameGrob) and populate it with plots and labels using a loop. It's useful for building multi-panel plots.

```R
fg <- frameGrob(layout=grid.layout(n.plots*2,
                    3,widths=unit(widths,"npc"),
                    heights=unit(c(rep(c(hit,hbet),n.plots-1),hit,hlast),"npc")))
  ## loop to create panels
  ## j indexes independent variables
  j <- 1
  for (i in index) {  ## i is the vertical slot position
    #create a dataframe with the data to plot now
    Y.now <- data.frame(estimate=Y$estimate[j,],
                        lo=Y$lo[j,],
                        hi=Y$hi[j,],
                        lo1=Y$lo1[j,],
                        hi1=Y$hi1[j,])
    ##similartly for Y2
    if (!is.null(Y2)) Y2.now <- data.frame(estimate=Y2$estimate[j,],
                                           lo=Y2$lo[j,],
                                           hi=Y2$hi[j,],
                                           lo1=Y2$lo1[j,],
                                           hi1=Y2$hi1[j,])
    else Y2.now <- NULL
    ## if it is the bottom row, set sp.now to a positive value
    if (i==max(index)) sp.now <- .1    
    ##are we drawing a reference line?
    drawRef <- !is.na(refline[j])
    ##place the plot
    ##the actual plot object is created by the function grid.Vdotplot
    fg <- placeGrob(fg,
                    grid.Vdotplot(Y.now,
                                  Y2.now,
                                  sp=c(.1,sp.now),draw=FALSE,lwd.fact=lwd.fact,
                                  refline=ifelse(drawRef,refline[j],0)) 
                    ,col=2,row=i)
    ##the independent variables labels
    fg <- placeGrob(fg,textGrob(x=pos.label.y,label.vec.vars[j]
                                ,rot=rot.label.y,gp=gpar(cex=.75*lwd.fact),just=just.label.y
                                ),col=1,row=i)
    j <- j+1
  }
  ## if Y2 exists and a legend is specified, draw it using the legendGrob function
  if (!is.null(Y2)&!is.null(legend)) {
    fg <- placeGrob(fg,legendGrob(c(21,21),legend,cex=leg.mult,leg.fontsize=leg.fontsize,fill=c("black","white")),col=1,row=i+1)
  }
  if (print) {
    grid.arrange(fg)
  } else {
    ## if we are not printing, return the graphical object
    fg
  }
```

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

### Format Data and Create Plot with dotwhisker

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

This snippet demonstrates how to format data into a tidy dataframe and then use the small_multiple function from the dotwhisker package to create a plot with custom themes and labels. It includes data preparation and plot customization steps.

```R
# Format data as tidy dataframe
model_names <- c("Full Sample\n",
                       "Excluding counties\nw/ partial registration\n",
                       "Full sample w/\nstate year dummies\n")
submodel_names <- c("With","Without")

model_order <- c(4, 1, 5, 2, 6, 3)

results_df <- data.frame(term = rep(varnames[1:6], times = 6),
                         estimate = as.vector(coef.matrix[, model_order]),
                         std.error = as.vector(se.matrix[, model_order]),
                         model = as.factor(rep(model_names, each = 12)),
                         submodel = rep(rep(submodel_names, each = 6), times = 3),
                         stringsAsFactors = FALSE)


small_multiple(results_df, show_stats = FALSE) + 
    scale_x_discrete(limits = model_names) + # order the models 
    theme_bw() + ylab("Coefficient Estimate") +
    geom_hline(yintercept = 0, colour = "grey60", linetype = 2) +
    theme(axis.text.x  = element_text(angle = 45, hjust = 1),
          legend.position=c(.97, .99), legend.justification=c(1, 1), 
          legend.title = element_text(size=8),
          legend.background = element_rect(color="gray90"),
          legend.spacing = unit(-3, "pt"),
          legend.key.size = unit(10, "pt")) + 
    scale_colour_hue(name = "Law Change\nDummy") +
    ggtitle("Registration Effects on Turnout\nAnsolabehere and Konisky (2006)\nvia Kastellec and Leoni (2007)")
```

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

### Basic dwplot Usage

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

Demonstrates the basic usage of the dwplot function with a pre-existing model data frame.

```r
dwplot(m1_df) #same as dwplot(m1)
```

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

### Rescaling Coefficients with by_2sd

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

Shows how to rescale coefficients by two standard deviations using the `by_2sd` function for easier comparison, particularly with dichotomous predictors. The results are then sorted alphabetically by term.

```R
# Customize the input data frame
m1_df_mod <-
    m1_df |> # the original tidy data.frame
    by_2sd(mtcars) |> # rescale the coefficients
    arrange(term) # alphabetize the variables

m1_df_mod # rescaled, with variables reordered alphabetically
```

```R
dwplot(m1_df_mod)
```

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

### Original R Code for Dot Plot with Error Bars

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

This code generates a dot plot with error bars for a single regression model using base R graphics. It requires manual setup of coefficient and standard error vectors, variable names, and plot margins.

```R
#Create vectors for coefficients, standard errors and variable names 
    #we place coefficient as last element in each vector rather than 1st
    #since it is least important predictor, and thus we place it at the bottom of the graph
#note: we exclude the constant, since it is substantively meaningless

coef.vec <- c( 1.31, .93, 1.46, .07, .96, .2, .22, -.21, -.32, -.27,.23, 
    0, -.03, .13, .15, .31, -.10)
se.vec <- c( .33, .32, .32, .37, .37, .13, .12, .12, .12, .07, .07, .01, .21,
    .14, .29, .25, .27)

var.names <- c("Argentina", "Chile", "Colombia", "Mexico", "Venezuela", #for longer names, we split into 2 lines using "\n" function
    "Retrospective Egocentric", "Prospective Egocentric",
    "Retrospective Sociotropic", "Prospective Sociotropic",
    "Distance from President", "Ideology", "Age", "Female", "Education",
    "Academic Sector", "Business Sector", "Government Sector")

    
y.axis <- c(length(coef.vec):1)#create indicator for y.axis, descending so that R orders vars from top to bottom on y-axis

oldpar <- par(mar=c(2, 13, 0, 0))#set margins for plot, leaving lots of room on left-margin (2nd number in margin command) for variable names
plot(coef.vec, y.axis, type = "p", axes = F, xlab = "", ylab = "", pch = 19, cex = 1.2,#plot coefficients as points, turning off axes and labels. 
    xlim = c(-2,2.5), xaxs = "r", main = "") #set limits of x-axis so that they include mins and maxs of 
        #coefficients + .95% confidence intervals and plot is symmetric; use "internal axes", and leave plot title empty
#the 3 lines below create horiztonal lines for 95% confidence intervals, and vertical ticks for 90% intervals
segments(coef.vec-qnorm(.975)*se.vec, y.axis, coef.vec+qnorm(.975)*se.vec, y.axis, lwd =  1.5)#coef +/-1.96*se = 95% interval, lwd adjusts line thickness

axis(1, at = seq(-2,2,by=.5), labels = NA, tick = T,#draw x-axis and labels with tick marks
    cex.axis = 1.2, mgp = c(2,.7,0))#reduce label size, moves labels closer to tick marks
axis(1, at = seq(-2,2,by=1), labels =  c(-2,  -1,  0, 1,2), tick = T,#draw x-axis and labels with tick marks
    cex.axis = 1.2, mgp = c(2,.7,0))#reduce label size, moves labels closer to tick marks    

axis(2, at = y.axis, label = var.names, las = 1, tick = T, mgp = c(2,.6,0),
    cex.axis = 1.2) #draw y-axis with tick marks, make labels perpendicular to axis and closer to axis
segments(0,0,0,17,lty=2) # draw dotted line through 0
#box(bty = "l") #place box around plot

#use following code to place model info into plot region
x.height <- .57
text(x.height, 10, expression(R^{2} == .15), adj = 0, cex = 1) #add text for R-squared
text(x.height, 9, expression(paste("Adjusted ", R^{2} == ".12", "")), adj = 0, cex = 1)#add text for Adjusted-R-squared
text(x.height, 8, "n = 500", adj = 0, cex = 1)#add text for sample size

```

```R
par(oldpar)
```

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

### Filtering with Regular Expressions

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

Illustrates filtering model estimates using regular expressions, useful for excluding factor levels like country dummies.

```r
# Transform cyl to factor variable in the data
m_factor <- lm(mpg ~ wt + cyl + disp + gear, data = mtcars |> mutate(cyl = factor(cyl)))

# Remove all model estimates that start with cyl*
m_factor_df <- broom::tidy(m_factor) |
 filter(!grepl('cyl*', term))

dwplot(m_factor_df)
```

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

### Create and Customize Regression Graph

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

This snippet shows how to create a regression graph using plot.reg and then customize its axis labels using editGrob. The graph is not printed until grid.arrange is called.

```R
tmp <- plot.reg(Y1,Y2,#the lists
                #the model labels
                label.x=c("Full Sample","Excluding counties\nw. partial registration",
                  "Full sample w. \nstate year dummies"),
                ## reference lines
                refline=c(0,0,0,0,0,0),
                ## space left in the bottom (for the x-axis labels)
                hlast=.15,
                ## print the graph?
                print=FALSE,
                ## line width / character size multiplier
                lwd.fact=1.3,
                ## length of the cross- hairs
                length.arrow=unit(0,"mm"),
                ## legend
                ##legend=c("without law change","with law change"),
                ## widths: variable names, plot size, y-axis
                widths=c(.6,.4,.3),
                ## rotation of the variable name labes
                rot.label.y=0,
                ## justification of the variable name labels
                just.label.y="right",
                ## position (x-axis) of the variable name labels)
                pos.label.y=0.95,
                ## size of the symbol
                pch.size=0.6,expand.factor=.2,expand.factor2=0.1,
                ##legend
                legend=c("With law\nchange dummy","Without law\nchange dummy"),leg.mult=.7,
                ##legend font size
                leg.fontsize=11, 
                v.grid=TRUE,
                yaxis.at=list(
                NULL,
                NULL,
                seq(-.1,.1,.05),
                seq(-.2,.2,.1),
                seq(-.2,.2,.1),
                NULL##seq(-1,1,.5)
                )
                )
tmp <- editGrob(tmp,gPath("xaxis","labels"),rot=45,just="right",gp=gpar(lineheight=.75))
##tmp is the object we have just created,"xaxis" is the name of element in the object with the x-axis
##elements, and "labels" is the actual object in xaxis that we want to rotate
##just is the justification of the text
grid.arrange(tmp) ## print the graph
```

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

### Extracting Coefficients for Unsupported Models

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

Demonstrates how to extract coefficients from model objects not directly supported by `tidy` or `parameters` using `coef()` and converting them into a data frame suitable for `dwplot`.

```R
# the ordinal regression model is not supported by tidy
m4 <- ordinal::clm(factor(gear) ~ wt + cyl + disp, data = mtcars)
m4_df <- coef(summary(m4)) |
    data.frame() |
    tibble::rownames_to_column("term") |
    rename(estimate = Estimate, std.error = Std..Error)
m4_df
```

```R
dwplot(m4_df)
```

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

### Display Model Fit Statistics in dwplot

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

Use `show_stats = TRUE` in `dwplot` to display model fit statistics below the plot. Adjust `stats_size` to control the text size of the statistics.

```r
dwplot(m1, show_stats = TRUE, stats_size = 3)
```

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

### Present Regression Results as Normal Distributions

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

Create dot-and-whisker plots styled as normal distributions by using `style = "distribution"` in `dwplot`. This allows for a visual representation of coefficient estimates and their uncertainty, similar to statistical distributions.

```R
by_transmission_brackets <- list(
    c("Overall", "Weight", "Weight"),
    c("Engine", "Cylinders", "Horsepower"),
    c("Transmission", "Gears", "1/4 Mile/t")
)

{
    mtcars %>%
        split(.$am) %>%
        purrr::map( ~ lm(mpg ~ wt + cyl + gear + qsec, data = .x)) %>%
        dwplot(style = "distribution") %>%
        relabel_predictors(
            wt = "Weight",
            cyl = "Cylinders",
            disp = "Displacement",
            hp = "Horsepower",
            gear = "Gears",
            qsec = "1/4 Mile/t"
        ) +
        theme_bw(base_size = 4) + xlab("Coefficient") + ylab("") +
        geom_vline(xintercept = 0,
                   colour = "grey60",
                   linetype = 2) +
        theme(
            legend.position = c(.995, .99),
            legend.justification = c(1, 1),
            legend.background = element_rect(colour = "grey80"),
            legend.title.align = .5
        ) +
        scale_colour_grey(
            start = .8,
            end = .4,
            name = "Transmission",
            breaks = c("Model 0", "Model 1"),
            labels = c("Automatic", "Manual")
        ) +
        scale_fill_grey(
            start = .8,
            end = .4,
            name = "Transmission",
            breaks = c("Model 0", "Model 1"),
            labels = c("Automatic", "Manual")
        ) +
        ggtitle("Predicting Gas Mileage by Transmission Type") +
    theme(plot.title = element_text(face = "bold", hjust = 0.5))
} |>
    add_brackets(by_transmission_brackets, fontSize = 0.3)
```

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

### Customizing dwplot Appearance with vline and Whisker Arguments

Source: https://fsolt.org/dotwhisker/news/index.html

Demonstrates how to customize the appearance of a dotwhisker plot by specifying colors for dots and whiskers, and by adding a custom vertical line behind the coefficients. This is useful for aesthetic preferences and highlighting specific values like zero.

```r
library(dotwhisker)
#> Loading required package: ggplot2

# linear model of interest
lm_object <- stats::lm(formula = wt ~ am * cyl, data = mtcars)

# creating the plot with dwplot
dwplot(x = lm_object,
       dot_args = list(color = "red"), # color for the dot
       whisker_args = list(color = "black"),   # color for the whisker
       vline = ggplot2::geom_vline(xintercept = 0,  # put vline _behind_ coefs
                                   colour = "grey60",
                                   linetype = 2,
                                   size = 1))

Created on 2018-06-27 by the reprex package (v0.2.0).
```

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

### Create Small Multiple Plot from List of Models

Source: https://fsolt.org/dotwhisker/reference/small_multiple.html

This snippet demonstrates how to generate a 'small multiple' plot by providing a list of `lm` model objects to the `small_multiple` function. Ensure your models are defined before passing them.

```r
m1 <- lm(mpg ~ wt + cyl + disp + gear, data = mtcars)
m2 <- update(m1, . ~ . + hp)


# Generate a 'small multiple' plot
small_multiple(list(m1, m2))
```

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

### Prepare Data for 'Secret Weapon' Plot

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

This R code prepares a dataset for creating 'secret weapon' plots by estimating linear models for subsets of data and tidying the results. It groups data by clarity and then applies `lm` and `broom::tidy` to generate model coefficients with confidence intervals.

```R
data(diamonds)

# Estimate models for many subsets of data, put results in a tidy data.frame
by_clarity <- diamonds %>%
    group_by(clarity) %>%
    do(broom::tidy(lm(price ~ carat + cut + color, data = .), conf.int = .99)) %>%
    ungroup() %>%
  rename(model = clarity)
```

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

### Create Coefficient and Standard Error Matrix

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

This R code snippet demonstrates how to create a matrix to hold coefficients and standard errors for multiple regression models. It's useful for preparing data for plotting or further analysis.

```r
#Create Vectors for coefs and standard errors for each model, and variable names
    #note that we  exclude "margin squared" since it doesn't appear in either model

    coef.matrix <- matrix(c(-.039, NA, .048, -.133, .071, -.795, 1.47, 
                        -.036, NA, .036, -.142, .07, -.834, 1.70, 
                        -.051, NA, .017, .05, .011, -.532, .775, 
                        -.037, -.02, .047, -.131,.072, -.783, 1.45,
                        -.034, -.018, -.035, -.139, .071, -.822, 1.68,
                        -.05, -.023, .016,-.049, .013, -.521, .819),nr=7)
```

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

### secret_weapon Function Signature

Source: https://fsolt.org/dotwhisker/reference/secret_weapon.html

This shows the function signature for `secret_weapon`, outlining its parameters and their default values.

```R
secret_weapon(x, var = NULL, ci = 0.95, margins = FALSE, by_2sd = FALSE, ...)
```

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

### Prepare Data for Dot-and-Whisker Plot

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

Formats regression results into a data frame suitable for the `dwplot` function. Ensure the data frame includes columns for term, estimate, standard error, and model.

```R
results_df <- data.frame(term = rep(var.names, times = 2),
                         estimate = c(coef.vec.1, coef.vec.2),
                         std.error = c(se.vec.1, se.vec.2),
                         model = c(rep("Model 1", 20), rep("Model 2", 20)),
                         stringsAsFactors = FALSE)

```

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

### Create and Customize Dot-and-Whisker Plot

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

Generates a basic dot-and-whisker plot and applies several customizations including theme, legend, labels, and a vertical line at zero. This is useful for comparing coefficients across models.

```R
p <- dwplot(results_df, show_stats = FALSE) +
    theme_bw() +
    theme(legend.justification=c(.02, .993), legend.position=c(.02, .99),
           legend.title = element_blank(), legend.background = 
              element_rect(color="gray90")) +
    xlab("Logit Coefficients") +
    geom_vline(xintercept = 0, colour = "grey60", linetype = 2) +
    ggtitle("Electoral Incentives and LDP Post Allocation\nPekkanen, Nyblade, and Krause (2006)\nvia Kastellec and Leoni (2007)")

```

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

### Filtering Model Estimates with tidy

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

Shows how to use the broom::tidy function to filter out specific model estimates, such as the intercept, before plotting.

```r
m1_df <- broom::tidy(m1) |> filter(term != "(Intercept)") |> mutate(model = "Model 1")
m2_df <- broom::tidy(m2) |> filter(term != "(Intercept)") |> mutate(model = "Model 2")

two_models <- rbind(m1_df, m2_df)

dwplot(two_models)
```

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

### Dotwhisker Plot for Regression Results

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

This code demonstrates how to use the dotwhisker package to create a plot of regression results. It requires formatting the data into a tidy dataframe with 'term', 'estimate', and 'std.error' columns.

```R
#install.packages("dotwhisker") # uncomment to install from CRAN
library(dplyr)
library(dotwhisker)
library(dplyr)

# Format data as tidy dataframe
results_df <- data.frame(term=var.names, estimate=coef.vec,
                               std.error=se.vec)

```

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

### Display Model Fit Statistics in small_multiple

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

Use `show_stats = TRUE` in `small_multiple` to display model fit statistics below each plot in a small multiples arrangement. `stats_size` adjusts the text size.

```r
small_multiple(list(m1, m2, m3), show_stats = TRUE, stats_size = 3)
```

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

### Plotting Model Coefficients and Confidence Intervals

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

Generates a plot comparing coefficient estimates and 95% confidence intervals for two models. Uses different plotting characters and colors to distinguish models. Includes manual legend creation.

```R
layout(matrix(c(2,1),1,2), #in order to add variable categories and braces to left side of plot, 
    widths = c(1.5, 5))#we use layout command, create a small second panel on left side.
                #using c(2,1) in matrix command tells R to create right panel 1st
#layout.show(2) #can use this command to check results of layout command (but it must be commented out when creating PDF).

oldpar <- par(mar=c(2,8,.5,1), lheight = .8)#set margins for regression plot
plot(coef.vec.1, y.axis+adjust, type = "p", axes = F, xlab = "", ylab = "", pch = 19, cex = 1.2, #plot model 1 coefs using black points (pch = 19, default = black), adding the "adjust amount" to the y.axis indicator to move points up
    xlim = c(min((coef.vec.1-qnorm(.975)*se.vec.1 -.1), (coef.vec.2-qnorm(.975)*se.vec.2 -.1), na.rm = T), #set xlims at mins and maximums (from both models) of confidence intervals, plus .1 to leave room at ends of plots
    max((coef.vec.1+qnorm(.975)*se.vec.1 -.1), (coef.vec.2+qnorm(.975)*se.vec.2 -.1), na.rm = T)),  #use na.rm=T since vectors have missing values
    ylim = c(min(y.axis), max(y.axis)), main = "")
axis(1,at = seq(-4,1, by = 1), label = seq(-4,1, by = 1), mgp = c(2,.8,1), cex.axis = 1.3)#add x-axis and labels; "pretty" creates a sequence of  equally spaced nice values that cover the range of the values in 'x'-- in this case, integers
axis(2, at = y.axis, label = var.names, las = 1, tick = T, cex.axis =1.3)#add y-axis and labels; las = 1 makes labels perpendicular to y-axis
#axis(3,pretty(coef.vec.1, 3))#same as x-axis, but on top axis 
abline(h = y.axis, lty = 2, lwd = .5, col = "light grey")#draw light dotted line at each variable for dotplot effect
#box(bty="l")#draw box around plot
segments(coef.vec.1-qnorm(.975)*se.vec.1, y.axis+adjust, coef.vec.1+qnorm(.975)*se.vec.1, y.axis+adjust, lwd =  1.3)#draw lines connecting 95% confidence intervals

abline(v=0, lty = 2) # draw dotted line through 0 for reference line for null significance hypothesis testing

#add 2nd model
    #because we are using white points and do want the lines to go "through" points rather than over them
        #we draw lines first and the overlay points
segments(coef.vec.2-qnorm(.975)*se.vec.2, y.axis-adjust, coef.vec.2+qnorm(.975)*se.vec.2, y.axis-adjust, lwd =  1.3)#draw lines connecting 95% confidence intervals
points(coef.vec.2, y.axis-adjust, pch = 21, cex = 1.2, bg = "white" ) #add point estimates for 2nd model; pch = 21 uses for overlay points, and "white" for white color

#add legend (manually) to identify which dots denote model 1 and which denote model 2
#legend(-4.5, 20, c("Model 1", "Model 2"), pch = c(19,21),bty = "n")
points(-4, 19.5, pch = 19, cex  = 1.2)
text(-3.7, 19.5, "Model 1", adj = 0,cex  = 1.2)#left-justify text using adj  = 0
points(-4, 18.5, pch = 21,cex  = 1.2)
text(-3.7, 18.5, "Model 2", adj = 0,cex  = 1.2)#left-justify text using adj  = 0
par(oldpar)

#Create Variable Categories and Braces to go in 2nd plot

oldpar <- par(mar=c(2,0,.5,0)) #set margins--- bottom (1st number) and top (3rd number) must be the same as in 1st plot
plot(seq(0,1,length=length(var.names)), y.axis, type = "n", axes = F,  xlab = "", ylab = "")#call empty plot using type="n"
    #use a sequence of length 20 so that x and y have same length

left.side <- .55#use this to manipulate how far segments are from y-axis

```

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

### BibTeX Citation for dotwhisker

Source: https://fsolt.org/dotwhisker/authors.html

Use this BibTeX entry to cite the dotwhisker R package in academic publications.

```bibtex
@Manual{
  title = {dotwhisker: Dot-and-Whisker Plots of Regression Results},
  author = {Frederick Solt and Yue Hu},
  year = {2024},
  note = {R package version 0.8.3},
  url = {https://fsolt.org/dotwhisker/},
}
```

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

### Creating a Tidy Data Frame of Regression Results

Source: https://fsolt.org/dotwhisker/articles/dotwhisker-vignette.html

Converts a regression model object into a tidy data frame using `broom::tidy()`. This data frame can then be used as input for `dwplot`.

```r
# regression compatible with tidy
m1_df <- broom::tidy(m1) # create data.frame of regression results
m1_df # a tidy data.frame available for dwplot
```

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

### Create a Graphical Tree (gTree) for Plotting

Source: https://fsolt.org/dotwhisker/articles/kl2007_examples.html

This function constructs a gTree object, which is a graphical object containing various components like grids, reference lines, confidence intervals, estimates, and axes. It defines the viewport for the plot area and can optionally draw the plot.

```R
gplot <- with(Y,
                ## gTree is a graphical object with the whole plot
                gTree(
                      children=gList(
                        hgrid,vgrid,
                        refline,
                        ci1a,
                        ci1b,
                        estimates.grob(x,Y),
                        ## if Y2
                        ci2a,
                        ci2b
                        ,points2
                        ## box/rectangle around the plot area
                        ##,rectGrob(gp=gpar(lwd=.5*lwd.fact))
                        ## plot x axis if sp2>0 (we name it xaxis, so we can refer to it later)
                        ,if(sp[2]!=0) xaxisGrob(at=x.o,label=label.x,name="xaxis",
                                gp=gpar(cex=0.8*lwd.fact,lwd=0.6*lwd.fact))
                        ## plot x axis with no labels if it is not the bottom plot
                        ,if((sp[2]==0)&(!v.grid)) xaxisGrob(at=x.o,label=rep("",length(x.o)),
                                gp=gpar(cex=0.8*lwd.fact,lwd=0.6*lwd.fact))
                        ## plot y-axis if sp1>0
                        ,if(sp[1]!=0) yaxisGrob(at=aty,label=labely,
                                gp=gpar(cex=0.7*lwd.fact,lwd=0.6*lwd.fact),main=FALSE,name="yaxis")),
                      ## definition of the viewport (plot area)
                      vp=viewport(width=unit(1, "npc"), 
                        height=unit(1, "npc"),
                        ##xscale=c(1,nrow(Y)),
                        xscale=c(1-.5,nrow(Y)+.5),
                        yscale=r.y
                        ##yscale=c(-.1,.1)
                        ,clip=FALSE)
                      ))
  if (draw==TRUE) {
    ##draw the plot
    grid.newpage()
    fg <- frameGrob(layout=grid.layout(2,2,widths=unit(c(sp[1],1-sp[1]),"npc"),heights=unit(c(1-sp[2],sp[2]),"npc")))
    fg <- placeGrob(fg,gplot,col=2,row=1)
    grid.arrange(fg)
  } else {
    gplot
  }
}
```
Tokens:
Raw