| Title: | Toolkit for Compiling, (Post-Hoc) Testing, and Plotting Regression Results |
|---|---|
| Description: | Compiling regression results into a publishable format, conducting post-hoc hypothesis testing, and plotting moderating effects (the effect of X on Y becomes stronger/weaker as Z increases). |
| Authors: | Rui K. Yang [aut, cre], Luyao Peng [aut] |
| Maintainer: | Rui K. Yang <[email protected]> |
| License: | GPL-3 |
| Version: | 0.1.3 |
| Built: | 2024-11-05 03:56:58 UTC |
| Source: | https://github.com/rkzyang/regrrr |
Add row numbers to regression result data.frame
add.n.r(df)add.n.r(df)
df |
a data.frame of regression result |
Add approximate p-value based on t score or z score, when sample size is large
add.pr(df, z.col = 3, p.already = FALSE)add.pr(df, z.col = 3, p.already = FALSE)
df |
a data.frame of regression result |
z.col |
the column number of t score or z score |
p.already |
whether the regression result already contains p.value |
Add significance level marks to the regression result
add.sig(df, Pr.col = 5)add.sig(df, Pr.col = 5)
df |
a data.frame of regression result, e.g. summary(a_lm_model)$coefficients |
Pr.col |
the column number of p.value |
quickly check correlation matrix, or the correlation between a particular X and all other vars could be useful for looking for relevant instrument
check_cor(data, var_name_select = NULL, d = 3)check_cor(data, var_name_select = NULL, d = 3)
data |
a data.frame used in regression model |
var_name_select |
to specify the variable names to be included in the table, default is NULL–all variables are included |
d |
number of digits retained after the decimal point |
data(mtcars) check_cor(mtcars)data(mtcars) check_cor(mtcars)
quickly check the proportion of NAs in each columns of a dataframe
check_na_in(data, true_total = FALSE)check_na_in(data, true_total = FALSE)
data |
a data.frame |
true_total |
FALSE to show the percentage, TRUE to show the true number of missing values |
data(mtcars) check_na_in(mtcars)data(mtcars) check_na_in(mtcars)
quickly check the vifs in a regression model; for checking multi-collinearity
check_vif(data)check_vif(data)
data |
a data.frame used in regression model |
data(mtcars) model <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) check_vif(data = model$model)data(mtcars) model <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) check_vif(data = model$model)
Combine regression results from different models by columns
combine_long_tab( tbl_1, tbl_2, tbl_3 = NULL, tbl_4 = NULL, tbl_5 = NULL, tbl_6 = NULL, tbl_7 = NULL, tbl_8 = NULL, tbl_9 = NULL, tbl_10 = NULL, tbl_11 = NULL, tbl_12 = NULL, tbl_13 = NULL, tbl_14 = NULL, tbl_15 = NULL, tbl_16 = NULL, tbl_17 = NULL, tbl_18 = NULL, tbl_19 = NULL, tbl_20 = NULL )combine_long_tab( tbl_1, tbl_2, tbl_3 = NULL, tbl_4 = NULL, tbl_5 = NULL, tbl_6 = NULL, tbl_7 = NULL, tbl_8 = NULL, tbl_9 = NULL, tbl_10 = NULL, tbl_11 = NULL, tbl_12 = NULL, tbl_13 = NULL, tbl_14 = NULL, tbl_15 = NULL, tbl_16 = NULL, tbl_17 = NULL, tbl_18 = NULL, tbl_19 = NULL, tbl_20 = NULL )
tbl_1 |
the 1st data.frame of regression result |
tbl_2 |
the 2nd data.frame of regression result |
tbl_3 |
the 3rd data.frame of regression result |
tbl_4 |
the 4th data.frame of regression result |
tbl_5 |
the 5th data.frame of regression result |
tbl_6 |
the 6th data.frame of regression result |
tbl_7 |
the 7th data.frame of regression result |
tbl_8 |
the 8th data.frame of regression result |
tbl_9 |
the 9th data.frame of regression result |
tbl_10 |
the 10th data.frame of regression result |
tbl_11 |
the 11th data.frame of regression result |
tbl_12 |
the 12th data.frame of regression result |
tbl_13 |
the 13th data.frame of regression result |
tbl_14 |
the 14th data.frame of regression result |
tbl_15 |
the 15th data.frame of regression result |
tbl_16 |
the 16th data.frame of regression result |
tbl_17 |
the 17th data.frame of regression result |
tbl_18 |
the 18th data.frame of regression result |
tbl_19 |
the 19th data.frame of regression result |
tbl_20 |
the 20th data.frame of regression result |
data(mtcars) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) m2 <- update(m1, . ~ .+ wt * vs) summary(m1) summary(m2) combine_long_tab(to_long_tab(summary(m1)$coef), to_long_tab(summary(m2)$coef))data(mtcars) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) m2 <- update(m1, . ~ .+ wt * vs) summary(m1) summary(m2) combine_long_tab(to_long_tab(summary(m1)$coef), to_long_tab(summary(m2)$coef))
Compare regression models, which is compatible with the reg.table output # updated 9/13/2018 #
compare_models( model1, model2, model3 = NULL, model4 = NULL, model5 = NULL, model6 = NULL, model7 = NULL, model8 = NULL, model9 = NULL, model10 = NULL, model11 = NULL, model12 = NULL, model13 = NULL, model14 = NULL, model15 = NULL, model16 = NULL, model17 = NULL, model18 = NULL, model19 = NULL, model20 = NULL, likelihood.only = FALSE, round.digit = 3, main.effect.only = NULL, intn.effect.only = NULL )compare_models( model1, model2, model3 = NULL, model4 = NULL, model5 = NULL, model6 = NULL, model7 = NULL, model8 = NULL, model9 = NULL, model10 = NULL, model11 = NULL, model12 = NULL, model13 = NULL, model14 = NULL, model15 = NULL, model16 = NULL, model17 = NULL, model18 = NULL, model19 = NULL, model20 = NULL, likelihood.only = FALSE, round.digit = 3, main.effect.only = NULL, intn.effect.only = NULL )
model1 |
the 1st regression model |
model2 |
the 2nd regression model |
model3 |
the 3rd regression model |
model4 |
the 4th regression model |
model5 |
the 5th regression model |
model6 |
the 6th regression model |
model7 |
the 7th regression model |
model8 |
the 8th regression model |
model9 |
the 9th regression model |
model10 |
the 10th regression model |
model11 |
the 11th regression model |
model12 |
the 12th regression model |
model13 |
the 13th regression model |
model14 |
the 14th regression model |
model15 |
the 15th regression model |
model16 |
the 16th regression model |
model17 |
the 17th regression model |
model18 |
the 18th regression model |
model19 |
the 19th regression model |
model20 |
the 20th regression model |
likelihood.only |
whether or not to output the likelihood |
round.digit |
number of decimal places to retain |
main.effect.only |
specify col number of alternative main-effect models, if any |
intn.effect.only |
specify col number of alternative moderator models, if any |
data(mtcars) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) m2 <- update(m1, . ~ .+ wt * vs) compare_models(m1, m2)data(mtcars) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) m2 <- update(m1, . ~ .+ wt * vs) compare_models(m1, m2)
make the correlation matrix from the data.frame used in regression
cor.table( data, data_to_combine = NULL, var_name_select = NULL, all.var.names = NULL, d = 2 )cor.table( data, data_to_combine = NULL, var_name_select = NULL, all.var.names = NULL, d = 2 )
data |
a data.frame used in regression model, e.g. model$model |
data_to_combine |
another data.frame used for regression model, e.g. when you have similar set of X's but different Y's |
var_name_select |
optional: to specify the variable names used in regression to be included in the correlation matrix |
all.var.names |
optional: to rename all variable names, a string vector |
d |
number of decimal places to retain |
data(mtcars) model <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) cor.table(data = model$model)data(mtcars) model <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) cor.table(data = model$model)
load multiple packages
load.pkgs(pkg_name_vec)load.pkgs(pkg_name_vec)
pkg_name_vec |
a string vector of package names |
## Not run: load.pkgs(c("dplyr", "car", "purrr")) ## End(Not run)## Not run: load.pkgs(c("dplyr", "car", "purrr")) ## End(Not run)
plotting the marginal effect of X on Y, with or without one or multiple interaction terms
plot_effect( reg.coef, data, model, by_color = FALSE, x_var.name = NULL, y_var.name = NULL, moderator.name = NULL, min_x = 0.001, max_x = 0.999, mdrt_quantile_05 = NULL, mdrt_quantile_50 = NULL, mdrt_quantile_95 = NULL, mod.n.sd = 1, confidence_interval = FALSE, v = NULL, CI_Ribbon = FALSE, title = NULL, xlab = "X_Var.name", ylab = "Y_Var.name", moderator.lab = "Moderator_name", mdrt.low.name = "Low", mdrt.mid.name = NULL, mdrt.high.name = "High", y.high.lim = NULL, y.low.lim = NULL, spline_labels = c("LHS", "RHS") )plot_effect( reg.coef, data, model, by_color = FALSE, x_var.name = NULL, y_var.name = NULL, moderator.name = NULL, min_x = 0.001, max_x = 0.999, mdrt_quantile_05 = NULL, mdrt_quantile_50 = NULL, mdrt_quantile_95 = NULL, mod.n.sd = 1, confidence_interval = FALSE, v = NULL, CI_Ribbon = FALSE, title = NULL, xlab = "X_Var.name", ylab = "Y_Var.name", moderator.lab = "Moderator_name", mdrt.low.name = "Low", mdrt.mid.name = NULL, mdrt.high.name = "High", y.high.lim = NULL, y.low.lim = NULL, spline_labels = c("LHS", "RHS") )
reg.coef |
a coefficient matrix of regression result, e.g. summary(lm_model)$coef |
data |
the data used in regression, a data frame |
model |
the model object, such as a "lm" object |
by_color |
plot interactions by colors, otherwise by line types |
x_var.name |
x name in the regression model, a string |
y_var.name |
y name in the regression model, a string |
moderator.name |
moderating variable name in the regression model, a string |
min_x |
the min of x scale, in percentile of x |
max_x |
the max of x scale, in percentile of x |
mdrt_quantile_05 |
set the low level of moderator, in percentile |
mdrt_quantile_50 |
set the middle level of moderator, in percentile |
mdrt_quantile_95 |
set the high level of moderator, in percentile |
mod.n.sd |
set the moderating strength, in the number of s.d. units, which can take negative values |
confidence_interval |
if TRUE, plot confidence intervals |
v |
a customized variance-covariance matrix |
CI_Ribbon |
if TRUE, plot confidence interval ribbons, if FALSE, plot error bars |
title |
the title of the plot |
xlab |
label of X |
ylab |
label of Y |
moderator.lab |
label of moderator |
mdrt.low.name |
the label of low-level moderator |
mdrt.mid.name |
the label of mid-level moderator |
mdrt.high.name |
the label of high-level moderator |
y.high.lim |
specify the upper limit of y |
y.low.lim |
specify the lower limit of y |
spline_labels |
label of the spline variable; when the main varaible is a linear spline and spline labels are supplied, the moderation effect will be presented by facets. |
## Not run: data(mtcars) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) plot_effect(reg.coef = summary(m1)$coefficients, data = mtcars, model = m1, x_var.name = "wt", y_var.name = "mpg", moderator.name = "hp", confidence_interval = TRUE, CI_Ribbon = TRUE, xlab = "Weight", ylab = "MPG", moderator.lab = "Horsepower") ## End(Not run) #' @examples ## Not run: data(mtcars) m2 <- lm(mpg ~ vs + carb + hp + wt + wt * hp + wt * vs, data = mtcars) plot_effect(reg.coef = summary(m2)$coefficients, data = mtcars, model = m2, x_var.name = "wt", y_var.name = "mpg", moderator.name = "hp", confidence_interval = TRUE, CI_Ribbon = FALSE, xlab = "Weight", ylab = "MPG", moderator.lab = "Horsepower") ## End(Not run) #' @examples ## Not run: data(mtcars) m3 <- lm(mpg ~ vs + carb + hp + lspline(wt, knots = 4, marginal = FALSE) * hp, data = mtcars) plot_effect(reg.coef=summary(m3)$coefficients, data = mtcars, model = m3, x_var.name = "wt", y_var.name = "mpg", moderator.name = "hp", xlab="Weight", ylab="MPG", moderator.lab="Horsepower") ## End(Not run) ## Not run: # this shows the function is compatible with ggplot2 customization library(extrafont) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) plot_effect(reg.coef = summary(m1)$coefficients, data = mtcars, model = m1, x_var.name = "wt", y_var.name = "mpg", moderator.name = "hp", confidence_interval = TRUE, CI_Ribbon = TRUE, xlab = "Weight", ylab = "MPG", moderator.lab = "Horsepower") + ggplot2::theme(text=ggplot2::element_text(family="Times New Roman", size = 16)) ## End(Not run)## Not run: data(mtcars) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) plot_effect(reg.coef = summary(m1)$coefficients, data = mtcars, model = m1, x_var.name = "wt", y_var.name = "mpg", moderator.name = "hp", confidence_interval = TRUE, CI_Ribbon = TRUE, xlab = "Weight", ylab = "MPG", moderator.lab = "Horsepower") ## End(Not run) #' @examples ## Not run: data(mtcars) m2 <- lm(mpg ~ vs + carb + hp + wt + wt * hp + wt * vs, data = mtcars) plot_effect(reg.coef = summary(m2)$coefficients, data = mtcars, model = m2, x_var.name = "wt", y_var.name = "mpg", moderator.name = "hp", confidence_interval = TRUE, CI_Ribbon = FALSE, xlab = "Weight", ylab = "MPG", moderator.lab = "Horsepower") ## End(Not run) #' @examples ## Not run: data(mtcars) m3 <- lm(mpg ~ vs + carb + hp + lspline(wt, knots = 4, marginal = FALSE) * hp, data = mtcars) plot_effect(reg.coef=summary(m3)$coefficients, data = mtcars, model = m3, x_var.name = "wt", y_var.name = "mpg", moderator.name = "hp", xlab="Weight", ylab="MPG", moderator.lab="Horsepower") ## End(Not run) ## Not run: # this shows the function is compatible with ggplot2 customization library(extrafont) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) plot_effect(reg.coef = summary(m1)$coefficients, data = mtcars, model = m1, x_var.name = "wt", y_var.name = "mpg", moderator.name = "hp", confidence_interval = TRUE, CI_Ribbon = TRUE, xlab = "Weight", ylab = "MPG", moderator.lab = "Horsepower") + ggplot2::theme(text=ggplot2::element_text(family="Times New Roman", size = 16)) ## End(Not run)
Compiling, Testing, Plotting Regression Results
Rui K. Yang [email protected]
Luyao Peng [email protected]
Useful links:
Report bugs at https://github.com/RkzYang/regrrr/issues
Scale a vector into the 0-1 scale
scale_01(x)scale_01(x)
x |
a vector |
testing equality of two coefficients (difference between coefficients of regressors), a Wald test note: if v is not alternatively specified, use car::linearHypothesis(lm_model, "X1 = X2")
test_coef_equality(model, var1.name, var2.name, v = NULL)test_coef_equality(model, var1.name, var2.name, v = NULL)
model |
the model object, such as a "lm" object |
var1.name |
X1 name in model, a string |
var2.name |
X2 name in model, a string |
v |
a customized variance-covariance matrix data(mtcars) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) summary(m1) test_coef_equality(model = m1, var1.name = "carb", var2.name = "hp") |
significance of regression slope (the marginal effect) under moderation testing restriction: the sig. of beta_x under the moderation of z1, with or without additional interaction terms (z2, z3, etc.)
test_tilted_slopes( reg.coef, v = NULL, model, x_var.name, moderator.name, mod.n.sd = 1, data, t.value.col = 3, Pr.col = 4 )test_tilted_slopes( reg.coef, v = NULL, model, x_var.name, moderator.name, mod.n.sd = 1, data, t.value.col = 3, Pr.col = 4 )
reg.coef |
a data.frame (or matrix) of regression result or a coeftest object, e.g. summary(lm_model)$coef, coeftest(lm_model, cluster.vcov(lm_model, cbind(data$group1, data$group2))) |
v |
a customized variance-covariance matrix |
model |
the model object, such as a "lm" object |
x_var.name |
main independent variable name in model, a string |
moderator.name |
moderator name in model, a string |
mod.n.sd |
specify the strength of the moderating effects, in the unit of s.d.s of the moderator, which can take negative values |
data |
data used for regression |
t.value.col |
col number of the t-score in reg.coef |
Pr.col |
col number of the Prob.(>|t|)) in reg.coef |
data(mtcars) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) test_tilted_slopes(reg.coef = summary(m1)$coef, model = m1, x_var.name = "wt", moderator.name = "hp", data = mtcars)data(mtcars) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) test_tilted_slopes(reg.coef = summary(m1)$coef, model = m1, x_var.name = "wt", moderator.name = "hp", data = mtcars)
Convert the regression result to the long format: the standard errors are in parentheses and beneath the betas
to_long_tab(reg.coef, d = 3, t.value.col = 3, Pr.col = 4)to_long_tab(reg.coef, d = 3, t.value.col = 3, Pr.col = 4)
reg.coef |
a data.frame (or matrix) of regression result or a coeftest object, e.g. summary(lm_model)$coef, coeftest(lm_model, cluster.vcov(lm_model, cbind(data$group1, data$group2))) |
d |
number of decimal places to retain |
t.value.col |
col number of the t-score in the reg.coef data.frame |
Pr.col |
col number of the Prob.(>|t|)) in the reg.coef data.frame |
data(mtcars) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) to_long_tab(reg.coef = summary(m1)$coef)data(mtcars) m1 <- lm(mpg ~ vs + carb + hp + wt + wt * hp , data = mtcars) to_long_tab(reg.coef = summary(m1)$coef)