hdpGLM (version 1.0.4)

plot_hdpglm: Plot posterior distributions

Description

this function creates a plot with two grids. One is the grid with posterior expectation of betas as function of context-level covariates. The other is the posterior distribution of tau

Usage

plot_hdpglm(
  samples,
  X = NULL,
  W = NULL,
  ncol.taus = 1,
  ncol.betas = NULL,
  ncol.w = NULL,
  nrow.w = NULL,
  smooth.line = FALSE,
  pred.pexp.beta = FALSE,
  title.tau = NULL,
  true.tau = NULL,
  title.beta = NULL,
  tau.x.axis.size = 1.1,
  tau.y.axis.size = 1.1,
  tau.title.size = 1.2,
  tau.panel.title.size = 1.4,
  tau.legend.size = 1,
  beta.x.axis.size = 1.1,
  beta.y.axis.size = 1.1,
  beta.title.size = 1.2,
  beta.panel.title.size = 1.4,
  beta.legend.size = 1,
  tau.xlab = NULL
)

Arguments

samples

an output of the function hdpGLM

X

a string vector with the name of the first-level covariates whose associated tau should be displayed

W

a string vector with the name of the context-level covariate(s) whose linear effect will be displayed. If NULL, the linear effect tau of all context-level covariates are displayed. Note: the context-level covariate must have been included in the estimation of the model.

ncol.taus

integer with the number of columns of the grid containing the posterior distribution of tau

ncol.betas

integer with the number of columns of the posterior expectation of betas as function of context-level features

ncol.w

integer with the number of columns to use to display the different context-level covariates

nrow.w

integer with the number of rows to use to display the different context-level covariates

smooth.line

boolean, if TRUE the plot will display a regression line representing the regression of the posterior expectation of the linear coefficients betas on the context-level covariates. Default FALSE

pred.pexp.beta

boolean, if TRUE the plots will display a line with the predicted posterior expectation of betas obtained using the posterior expectation of taus, the linear coefficients of the expectation of beta

title.tau

string, the title for the posterior distribution of the context effects

true.tau

a data.frame with four columns. The first must be named w and it indicates the index of each context-level covariate, starting with 0 for the intercept term. The second column named beta must contain the indexes of the betas of individual-level covariates, starting with 0 for the intercept term. The third column named Parameter must be named tau<w><beta>, where w and beta must be the actual values displayed in the columns w and beta. Finally, it must have a column named True with the true value of the parameter.

title.beta

string, the title for the posterior expectation of beta as function of context-level covariate

tau.x.axis.size

numeric, relative size of the x-axis of the plot with tau

tau.y.axis.size

numeric, relative size of the y-axis of the plot with tau

tau.title.size

numeric, relative size of the title of the plot with tau

tau.panel.title.size

numeric, relative size of the title of the panels of the plot with tau

tau.legend.size

numeric, relative size of the legend of the plot with tau

beta.x.axis.size

numeric, relative size of the x-axis of the plot with beta

beta.y.axis.size

numeric, relative size of the y-axis of the plot with beta

beta.title.size

numeric, relative size of the title of the plot with beta

beta.panel.title.size

numeric, relative size of the title of the panels of the plot with beta

beta.legend.size

numeric, relative size of the legend of the plot with beta

tau.xlab

string, the label of the x-axis for the plot with tau

Examples

library(magrittr)
# Note: this example is just for illustration. MCMC iterations are very reduced
set.seed(10)
n = 20
data.context1 = tibble::tibble(x1 = rnorm(n, -3),
                                   x2 = rnorm(n,  3),
                                   z  = sample(1:3, n, replace=TRUE),
                                   y  =I(z==1) * (3 + 4*x1 - x2 + rnorm(n)) +
                                       I(z==2) * (3 + 2*x1 + x2 + rnorm(n)) +
                                       I(z==3) * (3 - 4*x1 - x2 + rnorm(n)) ,
                                   w = 20
                                   ) 
data.context2 = tibble::tibble(x1 = rnorm(n, -3),
                                   x2 = rnorm(n,  3),
                                   z  = sample(1:2, n, replace=TRUE),
                                   y  =I(z==1) * (1 + 3*x1 - 2*x2 + rnorm(n)) +
                                       I(z==2) * (1 - 2*x1 +   x2 + rnorm(n)),
                                   w = 10
                                   ) 
data = data.context1 %>%
    dplyr::bind_rows(data.context2)

## estimation
mcmc    = list(burn.in=1, n.iter=50)
samples = hdpGLM(y ~ x1 + x2, y ~ w, data=data, mcmc=mcmc, n.display=1)

plot_hdpglm(samples)
plot_hdpglm(samples, ncol.taus=2, ncol.betas=2, X='x1')
plot_hdpglm(samples, ncol.taus=2, ncol.betas=2, X='x1', ncol.w=2, nrow.w=1,
            pred.pexp.beta=TRUE,smooth.line=TRUE )