updated pkg docs

README.Rmd

@@ -20,6 +20,7 @@ knitr::opts_chunk$set(
 ![Build Status](https://img.shields.io/badge/build-passing-brightgreen)
 ![Lifecycle:Stable](https://img.shields.io/badge/Lifecycle-Stable-97ca00)
 [![R-CMD-check](https://github.com/r-lib/rcmdcheck/actions/workflows/R-CMD-check.yaml/badge.svg)](https://github.com/r-lib/rcmdcheck/actions/workflows/R-CMD-check.yaml)
+[![CRAN Status](https://www.r-pkg.org/badges/version/RegrCoeffsExplorer)](https://CRAN.R-project.org/package=RegrCoeffsExplorer)
 <!-- badges: end -->
 
 _Always present effect sizes for primary outcomes_ [@Wilkinson1999Statistical].
@@ -40,7 +41,15 @@ Elastic-Net Regularized Generalized Linear Models (GLMNET) frameworks.
 
 ## Installation
 
-You can install the current version of `RegrCoeffsExplorer` from 
+CRAN version[@R-RegrCoeffsExplorer] can be installed with:
+
+```{r, eval=F, echo=T, results="hide", warning=F, message=F }
+
+install.packages("RegrCoeffsExplorer")
+
+```
+
+You can install the development version of `RegrCoeffsExplorer` from 
 [GitHub](https://github.com/vadimtyuryaev/RegrCoeffsExplorer) with:
 
 ```{r, eval=F, echo=T, results="hide", warning=F, message=F }
@@ -240,6 +249,8 @@ median and jitters to add random noise to data points preventing overlap and
 revealing the underlying data distribution more clearly. Substantial changes in
 the OR progressing alone the empirical data are clearly observed. 
 
+
+
 ### Customize plots - 1/2
 
 ```{r, warning=F, message=F}
@@ -319,24 +330,13 @@ how these interactions can be estimated and interpreted on the probability scale
 Consider a Linear Model with two continuous predictors and an interaction term:
 
 $$E[Y|\textbf{X}] = \beta_0 + \beta_1 x_1 + \beta_2 x_2 + \beta_{12} x_1 x_2$$
-
-Define the **marginal effect** by taking the partial derivative with respect to 
-$x_2$:
-
-$$\gamma_2 = \frac{\partial E[Y|\textbf{X}]}{\partial x_2} = \beta_2$$
-
-Therefore, $\beta_2$ is sufficient to quantify how much $E[Y|\textbf{X}]$ changes
-with respect to every one unit increase in $\beta_2$, holding all other variables
-constant. 
-
-Now, take the second order cross-partial derivative of $E[Y|\textbf{X}]$ with 
+Take the second order cross-partial derivative of $E[Y|\textbf{X}]$ with 
 respect to both $x_1$ and $x_2$:
 
 $$\gamma_{12}^2 = \frac{\partial^2 E[Y| \textbf{X}]}{\partial x_1 \partial x_2} = \beta_{12}$$
 
-Similar intuition as above holds. The interaction term $\beta_{12}$ shows 
-how effect of $x_1$ on $E[Y|\textbf{X}]$ changes for every one unit increase in
-$x_2$ and vice versa. 
+The interaction term $\beta_{12}$ shows how effect of $x_1$ on $E[Y|\textbf{X}]$ 
+changes for every one unit increase in $x_2$ and vice versa. 
 
 Now consider a logistic regression model with a non-linear link function $g(\cdot)$,
 two continuous predictors and an interaction term: 

README.md

@@ -8,6 +8,8 @@
 ![Build Status](https://img.shields.io/badge/build-passing-brightgreen)
 ![Lifecycle:Stable](https://img.shields.io/badge/Lifecycle-Stable-97ca00)
 [![R-CMD-check](https://github.com/r-lib/rcmdcheck/actions/workflows/R-CMD-check.yaml/badge.svg)](https://github.com/r-lib/rcmdcheck/actions/workflows/R-CMD-check.yaml)
+[![CRAN
+Status](https://www.r-pkg.org/badges/version/RegrCoeffsExplorer)](https://CRAN.R-project.org/package=RegrCoeffsExplorer)
 <!-- badges: end -->
 
 *Always present effect sizes for primary outcomes* (Wilkinson 1999).
@@ -33,7 +35,14 @@ Elastic-Net Regularized Generalized Linear Models (GLMNET) frameworks.
 
 ## Installation
 
-You can install the current version of `RegrCoeffsExplorer` from
+CRAN version(Tyuryaev et al. 2024) can be installed with:
+
+``` r
+
+install.packages("RegrCoeffsExplorer")
+```
+
+You can install the development version of `RegrCoeffsExplorer` from
 [GitHub](https://github.com/vadimtyuryaev/RegrCoeffsExplorer) with:
 
 ``` r
@@ -286,7 +295,7 @@ vis_reg(glm_model, CI = TRUE, intercept = TRUE,
   theme(plot.title = element_text(hjust = 0.5))   
 ```
 
-<img src="man/figures/README-unnamed-chunk-3-1.png" width="100%" /> As
+<img src="man/figures/README-unnamed-chunk-4-1.png" width="100%" /> As
 observed, when returning individual plots, the resulting entities are
 `ggplot` objects. Consequently, any operation that is compatible with
 ggplot can be applied to these plots using the `+` operator.
@@ -301,7 +310,7 @@ vis_reg(glm_model, CI = TRUE, intercept = TRUE,
              linetype="dashed", color = "orange", size=1)
 ```
 
-<img src="man/figures/README-unnamed-chunk-4-1.png" width="100%" />
+<img src="man/figures/README-unnamed-chunk-5-1.png" width="100%" />
 
 ## Vignettes
 
@@ -357,24 +366,14 @@ Consider a Linear Model with two continuous predictors and an
 interaction term:
 
 $$E[Y|\textbf{X}] = \beta_0 + \beta_1 x_1 + \beta_2 x_2 + \beta_{12} x_1 x_2$$
-
-Define the **marginal effect** by taking the partial derivative with
-respect to $x_2$:
-
-$$\gamma_2 = \frac{\partial E[Y|\textbf{X}]}{\partial x_2} = \beta_2$$
-
-Therefore, $\beta_2$ is sufficient to quantify how much
-$E[Y|\textbf{X}]$ changes with respect to every one unit increase in
-$\beta_2$, holding all other variables constant.
-
-Now, take the second order cross-partial derivative of $E[Y|\textbf{X}]$
-with respect to both $x_1$ and $x_2$:
+Take the second order cross-partial derivative of $E[Y|\textbf{X}]$ with
+respect to both $x_1$ and $x_2$:
 
 $$\gamma_{12}^2 = \frac{\partial^2 E[Y| \textbf{X}]}{\partial x_1 \partial x_2} = \beta_{12}$$
 
-Similar intuition as above holds. The interaction term $\beta_{12}$
-shows how effect of $x_1$ on $E[Y|\textbf{X}]$ changes for every one
-unit increase in $x_2$ and vice versa.
+The interaction term $\beta_{12}$ shows how effect of $x_1$ on
+$E[Y|\textbf{X}]$ changes for every one unit increase in $x_2$ and vice
+versa.
 
 Now consider a logistic regression model with a non-linear link function
 $g(\cdot)$, two continuous predictors and an interaction term:
@@ -583,7 +582,7 @@ ggplot(long_gamma_df, aes(x = X2_Quantile, y = GammaSquared)) +
   theme(plot.title = element_text(hjust = 0.5))
 ```
 
-<img src="man/figures/README-unnamed-chunk-7-1.png" width="100%" />
+<img src="man/figures/README-unnamed-chunk-8-1.png" width="100%" />
 
 Note that the estimate of the interaction term is positive ($0.68345$).
 Yet, significant number of the gamma squared values are negative.
@@ -651,7 +650,7 @@ lines(x1_values, predictions[[3]], lty = 3, lwd = 2)
 legend("topleft", legend = c("25% X2b", "50% X2b", "75% X2b"), lty = 1:3, lwd = 2)
 ```
 
-<img src="man/figures/README-unnamed-chunk-8-1.png" width="100%" />
+<img src="man/figures/README-unnamed-chunk-9-1.png" width="100%" />
 
 The alterations in $\hat{E}[Y|\textbf{x}]$ associated with one-unit
 increments in `X1b` at the first and third quartiles of `X2b`
@@ -733,6 +732,16 @@ Research:apantheon of Statistical Significance and Other Faux Pas.”
 
 </div>
 
+<div id="ref-R-RegrCoeffsExplorer" class="csl-entry">
+
+Tyuryaev, Vadim, Aleksandr Tsybakin, Jane Heffernan, Hanna Jankowski,
+and Kevin McGregor. 2024. *RegrCoeffsExplorer: Efficient Visualization
+of Regression Coefficients for Lm(), Glm(), and Glmnet() Objects*.
+Comprehensive R Archive Network (CRAN).
+<https://doi.org/10.32614/CRAN.package.RegrCoeffsExplorer>.
+
+</div>
+
 <div id="ref-Wilkinson1999Statistical" class="csl-entry">
 
 Wilkinson, Leland. 1999. “Statistical Methods in Psychology Journals:

references.bib

@@ -73,3 +73,16 @@ @book{Hastie2015
 publisher = {Chapman \& Hall/CRC},
 url={https://www.ime.unicamp.br/~dias/SLS.pdf}}
 
+@Manual{R-RegrCoeffsExplorer,
+  title        = {RegrCoeffsExplorer: Efficient Visualization of Regression Coefficients for lm(), glm(), and glmnet() Objects},
+  author       = {Vadim Tyuryaev and Aleksandr Tsybakin and Jane Heffernan and Hanna Jankowski and Kevin McGregor},
+  year         = {2024},
+  note         = {R package version 1.1.0},
+  doi          = {10.32614/CRAN.package.RegrCoeffsExplorer},
+  url          = {https://CRAN.R-project.org/package=RegrCoeffsExplorer},
+  abstract     = {The visualization tool offers a nuanced understanding of regression dynamics, going beyond traditional per-unit interpretation of continuous variables versus categorical ones. It highlights the impact of unit changes as well as larger shifts like interquartile changes, acknowledging the distribution of empirical data. Furthermore, it generates visualizations depicting alterations in Odds Ratios for predictors across minimum, first quartile, median, third quartile, and maximum values, aiding in comprehending predictor-outcome interplay within empirical data distributions, particularly in logistic regression frameworks.},
+  publisher    = {Comprehensive R Archive Network (CRAN)},
+  maintainer   = {Vadim Tyuryaev <vadim.tyuryaev@gmail.com>},
+  license      = {MIT + file LICENSE}
+}
+

vignettes/BetaVisualizer.Rmd

@@ -296,7 +296,7 @@ their interpretation align with the paradigm change discussed previously.
 plt_1=vis_reg(best_model_lasso,eff_size_diff=c(1,3),
               glmnet_fct_var="Originnon-USA")$"PerUnitVis"+
               ggtitle("Visualization of CV.GLMNET Results (per unit change)")+
-              ylim(-4,4)+
+              ylim(-5,5)+
               xlab("Car characteristics")+
               ylab("LASSO coefficients")+
               theme_bw()+
@@ -368,12 +368,17 @@ heart_cont_table
 # 'age' is skewed variable with a very big range
 paste("Range of '\ age \' variable is : ",diff(range(heart_df_filtered$age)))
 
-old_par = par()      
+# Save the current graphical parameters
+old_par = par(no.readonly = TRUE) 
+
+# Produce boxplots and histograms 
 par(mfrow=c(2,2))
 hist(heart_df_filtered$age, main="Histogram of Age", xlab="age")
 boxplot(heart_df_filtered$age,main="Boxplot of Age", ylab="age")
 hist(sqrt(heart_df_filtered$age),main="Histogram of transformed data", xlab="Sqrt(age)")
 boxplot(sqrt(heart_df_filtered$age),main="Boxplot of transformed data", ylab="Sqrt(age)")
+
+# Restore previous graphical parameters
 par(old_par)          
 
 ```
@@ -495,7 +500,7 @@ or up before passing the data to the function.
 
 ### 3.6.6 LASSO regression with CIs and custom realized effect size
 
-```{r,fig.height=6,fig.width=12, warning=F}
+```{r,fig.height=6,fig.width=12, warning=T}
 
 grid.arrange(vis_reg(out_heart, CI=T, glmnet_fct_var=c("surgery1"), 
                      round_func="none",eff_size_diff=c(1,3))$"SidebySide"