Conduct a floodlight analysis for a set of contrasts with a continuous moderator variable.
floodlight_for_contrasts(
data = NULL,
iv_name = NULL,
dv_name = NULL,
mod_name = NULL,
contrasts = NULL,
contrasts_for_floodlight = NULL,
covariate_name = NULL,
interaction_p_include = TRUE,
iv_category_order = NULL,
heteroskedasticity_consistent_se = "HC4",
round_r_squared = 3,
round_f = 2,
sigfigs = 2,
jn_points_disregard_threshold = NULL,
print_floodlight_plots = TRUE,
output = "reg_lines_plot",
jitter_x_percent = 0,
jitter_y_percent = 0,
dot_alpha = 0.5,
dot_size = 4,
interaction_p_value_font_size = 6,
jn_point_font_size = 6,
jn_point_label_hjust = NULL,
interaction_p_vjust = -3,
plot_margin = ggplot2::unit(c(75, 7, 7, 7), "pt"),
legend_position = "right",
line_of_fit_types = c("solid", "dashed"),
line_of_fit_thickness = 1.5,
jn_line_types = c("solid", "solid"),
jn_line_thickness = 1.5,
sig_region_color = "green",
sig_region_alpha = 0.08,
nonsig_region_color = "gray",
nonsig_region_alpha = 0.08,
x_axis_title = NULL,
y_axis_title = NULL,
legend_title = NULL,
round_decimals_int_p_value = 3,
round_jn_point_labels = 2
)a data object (a data frame or a data.table)
name of the multicategorical independent variable; this variable must have three or more categories.
name of the dependent variable
name of the continuous moderator variable
names of the contrast variables
names of the contrast variables for which floodlight analyses will be conducted
name of the variables to control for
logical. Should the plot include a p-value for the interaction term?
order of levels in the independent
variable for legend. By default, it will be set as levels of the
independent variable ordered using R's base function sort.
which kind of heteroskedasticity-consistent (robust) standard errors should be calculated? (default = "HC4")
number of decimal places to which to round r-squared values (default = 3)
number of decimal places to which to round the f statistic for model comparison (default = 2)
number of significant digits to round to
(for values in the regression tables, except for p values).
By default sigfigs = 2
the Minimum Distance in
the unit of the moderator variable that will be used for various purposes,
such as (1) to disregard the second Johnson-Neyman point
that is different from the first Johnson-Neyman (JN) point by
less than the Minimum Distance; (2) to determine regions of
significance, which will calculate the p-value of the IV's effect
(the focal dummy variable's effect) on DV at a candidate
JN point + / - the Minimum Distance.
This input is hard to explain, but a user can enter a really low value
for this argument (e.g., jn_points_disregard_threshold = 0.1
for a moderator measured on a 100-point scale) or use the default.
By default,
jn_points_disregard_threshold = range of the moderator / 10000
For example, if the observed moderator values range from 1 to 7
(because it is a 7-point scale), then
jn_points_disregard_threshold = (7 - 1) / 10000 = 0.0006
If print_floodlight_plots = TRUE,
a floodlight plot for each dummy variable will be printed.
By default, print_floodlight_plots = TRUE
output of the function (default = "all"). Possible inputs: "reg_models", "reg_tables", "reg_tables_rounded", "all"
horizontally jitter dots by a percentage of the range of x values
vertically jitter dots by a percentage of the range of y values
opacity of the dots (0 = completely transparent,
1 = completely opaque). By default, dot_alpha = 0.5
size of the dots (default = 4)
font size for the interaction p value (default = 8)
font size for Johnson-Neyman point labels (default = 6)
a vector of hjust values for Johnson-Neyman point labels. By default, the hjust value will be 0.5 for all the points.
By how much should the label for the
interaction p-value be adjusted vertically?
By default, interaction_p_vjust = -3)
margin for the plot
By default plot_margin = ggplot2::unit(c(75, 7, 7, 7), "pt")
position of the legend (default = "right").
If legend_position = "none", the legend will be removed.
types of the lines of fit for the two levels
of the independent variable.
By default, line_of_fit_types = c("solid", "dashed")
thickness of the lines of fit (default = 1.5)
types of the lines for Johnson-Neyman points.
By default, jn_line_types = c("solid", "solid")
thickness of the lines at Johnson-Neyman points (default = 1.5)
color of the significant region, i.e., range(s) of the moderator variable for which simple effect of the independent variable on the dependent variable is statistically significant.
opacity for sig_region_color.
(0 = completely transparent, 1 = completely opaque).
By default, sig_region_alpha = 0.08
color of the non-significant region, i.e., range(s) of the moderator variable for which simple effect of the independent variable on the dependent variable is not statistically significant.
opacity for nonsig_region_color.
(0 = completely transparent, 1 = completely opaque).
By default, nonsig_region_alpha = 0.08
title of the x axis. By default, it will be set
as input for mod_name. If x_axis_title = FALSE, it will
be removed.
title of the y axis. By default, it will be set
as input for dv_name. If y_axis_title = FALSE, it will
be removed.
title of the legend. By default, it will be set
as input for iv_name. If legend_title = FALSE, it will
be removed.
To how many digits after the decimal point should the p value for the interaction term be rounded? (default = 3)
To how many digits after the decimal point should the jn point labels be rounded? (default = 2)
See the following reference, which covers a related topic: Hayes & Montoya (2017) tools:::Rd_expr_doi("10.1080/19312458.2016.1271116")
if (FALSE) {
# typical example
# copy and modify the 'mtcars' data
mtcars2 <- setDT(data.table::copy(mtcars))
# make sure the data table package is attached
mtcars2[, contrast_1 := fcase(cyl == 4, -2, cyl %in% c(6, 8), 1)]
mtcars2[, contrast_2 := fcase(cyl == 4, 0, cyl == 6, 1, cyl == 8, -1)]
floodlight_for_contrasts(
data = mtcars2,
iv_name = "cyl",
dv_name = "mpg",
mod_name = "qsec",
contrasts = paste0("contrast_", 1:2),
contrasts_for_floodlight = "contrast_2")
}
Run the code above in your browser using DataLab