plot.emdi {povmap}

Plots for an emdi Object

Description

Diagnostic plots of the underlying model in the EBP (see also ebp) or Fay-Herriot (see also fh) approaches are obtained. These include Q-Q plots and density plots of residuals and random effects from the nested error linear regression model/ the Fay-Herriot model, a Cook's distance plot for detecting outliers and the log-likelihood of the estimation of the optimal parameter in Box-Cox transformations (the latter two only for ebp). The return depends on the transformation such that a plot for the optimal parameter is only returned in case a transformation with transformation parameter is chosen. The range of the x-axis is optional but necessary to change if there are convergence problems. All plots are obtained by ggplot.

Usage

## S3 method for class 'emdi'
plot(
  x,
  label = "orig",
  color = c("blue", "lightblue3"),
  gg_theme = NULL,
  cooks = TRUE,
  range = NULL,
  ...
)

## S3 method for class 'ebp'
plot(
  x,
  label = "orig",
  color = c("blue", "lightblue3"),
  gg_theme = NULL,
  cooks = TRUE,
  range = NULL,
  ...
)

## S3 method for class 'direct'
plot(x, ...)

## S3 method for class 'fh'
plot(
  x,
  label = "orig",
  color = c("blue", "lightblue3"),
  gg_theme = NULL,
  cooks = TRUE,
  range = NULL,
  ...
)

Arguments

x

an object of type "emdi", either "ebp" or "fh", representing point and, if chosen, MSE estimates obtained by the EBP or Fay-Herriot approach (see also ebp and fh).

label

argument that enables to customize title and axis labels. There are three instant options to label the diagnostic plot: (i) original labels ("orig"), (ii) axis labels but no title ("no_title"), (iii) neither axis labels nor title ("blank"). (iv) individual labels by a list that needs to have below structure. Six elements can be defined called qq_res, qq_ran, d_res, d_ran, cooks and opt_lambda for the six different plots and these list elements need to have three elements each called title, y_lab and x_lab. Only the labels for the plots that should be different to the original need to be specified. Please see the details section for an example with the default labels.

color

a character vector with two elements. The first element defines the color for the line in the QQ-plots, for the Cook's Distance plot and for the Box-Cox plot. The second element defines the color for the densities.

gg_theme

theme list from package ggplot2. For using this argument, package ggplot2 must be loaded via library(ggplot2). See also Example 4.

cooks

if TRUE, a Cook's distance plot is returned when the ebp function is used. The used method mdffits.default from the package HLMdiag struggles when data sets get large. In these cases, cooks should be set to FALSE. It defaults to TRUE.

range

optional sequence determining the range of the x-axis for plots of the optimal transformation parameter that defaults to NULL. In that case a range of the default interval is used for the plots of the optimal parameter. This leads in some cases to convergence problems such that it should be changed to e.g. the selected interval. The default value depends on the chosen data driven transformation and equals the default interval for the estimation of the optimal parameter.

...

optional arguments passed to generic function.

Details

The default settings of the label argument are as follows (please note that the title for opt_lambda depends on the chosen transformation, for the example Box-Cox is shown):

list(

qq_res =

c(title="Error term", y_lab="Quantiles of pearson residuals", x_lab="Theoretical quantiles"),

qq_ran =

c(title="Random effect", y_lab="Quantiles of random effects", x_lab="Theoretical quantiles"),

d_res =

c(title="Density - Pearson residuals", y_lab="Density", x_lab="Pearson residuals"),

d_ran =

c(title="Density - Standardized random effects", y_lab="Density", x_lab="Standardized random effects"),

cooks =

c(title="Cook's Distance Plot", y_lab="Cook's Distance", x_lab="Index"),

opt_lambda =

c(title="Box-Cox - REML", y_lab="Log-Likelihood", x_lab="expression(lambda)"))

Value

Two Q-Q plots in one grid, two density plots, a Cook's distance plot and a likelihood plot for the optimal parameter of transformations with transformation parameter obtained by ggplot. The latter two plots are only provided for ebp object.

See Also

emdiObject, ebp, fh

Examples

# Examples for models of type ebp
# Loading data - population and sample data
data("eusilcA_pop")
data("eusilcA_smp")

# With default setting but na.rm = TRUE; with Box-Cox transformation
emdi_model <- ebp(
  fixed = eqIncome ~ gender + eqsize + cash + self_empl +
    unempl_ben + age_ben + surv_ben + sick_ben + dis_ben + rent + fam_allow +
    house_allow + cap_inv + tax_adj, pop_data = eusilcA_pop,
  pop_domains = "district", smp_data = eusilcA_smp, smp_domains = "district",
  na.rm = TRUE
)

# Example 1: Creation of default diagnostic plots
plot(emdi_model)

# Example 2: Creation of diagnostic plots without labels and titles,
# different colors and without Cook's distance plot.
plot(emdi_model,
  label = "no_title", color = c("red", "yellow"),
  cooks = FALSE
)

# Example 3: Creation of diagnostic plots where labels and title differs for
# residual plot
plot(emdi_model,
  label = list(qq_res = c(
    title = "Pearson resid.",
    y_lab = "Quant.", x_lab = "Theo. Quant."
  )), color = c("red", "yellow"),
  cooks = FALSE
)

# Example 4: Usage of theme from ggplot2 within plot.emdi
library(ggplot2)
plot(emdi_model, gg_theme = theme(
  panel.background =
    element_rect(fill = "white", colour = "white"),
  plot.title = element_text(face = "bold"),
  title = element_text(color = "navy")
))

# Example for models of type fh

# Loading data - population and sample data
data("eusilcA_popAgg")
data("eusilcA_smpAgg")

# Combine sample and population data
combined_data <- combine_data(
  pop_data = eusilcA_popAgg,
  pop_domains = "Domain",
  smp_data = eusilcA_smpAgg,
  smp_domains = "Domain"
)

# Generation of the emdi object
fh_std <- fh(
  fixed = Mean ~ cash + self_empl, vardir = "Var_Mean",
  combined_data = combined_data, domains = "Domain",
  method = "ml", MSE = TRUE
)

# Example 5: Creation of default diagnostic plots for Fay-Herriot model
plot(fh_std)