powerPlot: Power plots with ggplot2

In lfpdroubi/appraiseR: R Package For Real Estate Appraisals

Power plots with ggplot2

Description

powerPlot generates Power Plots for lm or bestfit objects with ggplot2.

Usage

powerPlot(y, yhat, ...)

## Default S3 method:
powerPlot(
  y,
  yhat,
  system = c("base", "ggplot2", "lattice"),
  axis = c("standard", "inverted"),
  Smooth = FALSE,
  methods = "lm",
  se = FALSE,
  metrics = c("none", "rmse", "mae", "mape"),
  R2 = FALSE,
  ...
)

## S3 method for class 'lm'
powerPlot(object, FUN, system = c("base", "ggplot2", "lattice"), ...)

## S3 method for class 'bestfit'
powerPlot(object, fit = 1, scale = c("transformed", "original"), ...)

## S3 method for class 'lmerMod'
powerPlot(object, FUN, ...)

## S3 method for class 'rq'
powerPlot(object, FUN, ...)

Arguments

y	observed values
yhat	predicted values
...	further args passed to powerPlot.default
axis	option to plot predicted values on the x axis (inverted) or in the y axis (standard)
Smooth	option to add a regression line to the plot
se	option to add confidence interval of regression line to the plot
metrics	TRUE or FALSE. If TRUE, metrics (RMSE, MAE and MAPE) are displayed at the plot area.
R2	TRUE or FALSE. If TRUE, R2 between predicted and observed values are printed.
object	object of class lm, rq, lmerMod or bestfit.
FUN	function used to transform the dependent variable (the inverse of this function will be used to retransform the data to the original scale).
fit	chosen fit
scale	Plot values in original or transformed scale?

Value

a power plot

Examples

library(sf)
library(broom)
library(ggplot2)
centro_2015$padrao <- as.numeric(centro_2015$padrao)
fit <- lm(log(valor) ~ area_total + quartos + suites + garagens +
          log(dist_b_mar) + I(1/padrao),
          data = centro_2015, subset = -c(31, 39))
s <- summary(fit)
centro_2015 <- expand.model.frame(fit, ~ valor + dist_b_mar + padrao,
                                  na.expand = T)
# Transformed Scale:
y <- centro_2015$`log(valor)`
powerPlot(y = y, yhat = fitted(fit),
          system = "ggplot2", Smooth = TRUE)
powerPlot(y = y, yhat = fitted(fit),
          system = "ggplot2", Smooth = TRUE, axis = "inverted")
# Original Scale:
## Mediana
Y <- centro_2015$valor
YHAT <-  exp(fitted(fit))
p <- powerPlot(y = Y,
               yhat = YHAT,
               system = 'ggplot2',
               axis = "inverted",
               Smooth = TRUE)
p + labs(title = "Poder de Predição", subtitle = "Mediana")
## Média
YHAT1 <- exp(fitted(fit) + s$sigma^2/2)
p1 <- powerPlot(y = Y,
                yhat = YHAT1,
                system = 'ggplot2',
                axis = "inverted",
                Smooth = TRUE)
p1 + labs(title = "Poder de Predição", subtitle = "Média")
## Moda
YHAT2 <- exp(fitted(fit) - s$sigma^2)
p2 <- powerPlot(y = Y,
                yhat = YHAT2,
                system = 'ggplot2',
                axis = "inverted",
                Smooth = TRUE)
p2 + labs(title = "Poder de Predição", subtitle = "Moda")
# powerPlot.lm:
par(pty="s") #default par(pty="m")
powerPlot(fit)
powerPlot(fit, main = "Poder de Predição")
powerPlot(fit, FUN = "log",
          main = "Poder de Predição")
title(sub = "Em reais", adj = 1)
powerPlot(fit, axis = "inverted")
p <- powerPlot(fit, system = "ggplot2",
               FUN = "log", axis = "inverted")
p +
  labs(title = "Poder de Predição", subtitle = "Em Reais") +
  theme(aspect.ratio = 1)
# Changes Dep Var Transformation
fit1 <- lm(sqrt(valor) ~ area_total + quartos + suites + garagens +
          log(dist_b_mar) + I(1/padrao), centro_2015)
powerPlot(fit1)
powerPlot(fit1, FUN = "sqrt", axis = "inverted")
# powerPlot.bestfit:
library(sf)
dados <- st_drop_geometry(centro_2015)
dados$padrao <- as.numeric(dados$padrao)
best_fit <- bestfit(valor ~ ., dados,
                    transf = c("rec", "rsqrt", "log", "sqrt"),
                    subset = -c(31, 39))
p1 <- powerPlot(best_fit, system ="ggplot2")
p2 <- powerPlot(best_fit, system ="ggplot2", scale = "original")
cowplot::plot_grid(p1, p2, nrow = 1, ncol = 2)
p1 <- powerPlot(best_fit, system = "ggplot2",
                axis = "inverted")
p2 <- powerPlot(best_fit, system = "ggplot2",
                axis = "inverted", scale = "original")
cowplot::plot_grid(p1, p2, nrow = 1, ncol = 2)
powerPlot(best_fit, fit = 514)
p1 <- powerPlot(best_fit, fit = 514, system = "ggplot2",
                axis = "inverted")
p2 <- powerPlot(best_fit, fit = 514, system = "ggplot2",
                scale = "original", axis = "inverted")
cowplot::plot_grid(p1, p2, nrow = 1, ncol = 2)

# powerPlot.lmerMod:
library(lme4)
Mfit <- lmer(log(valor) ~ area_total + quartos + suites + garagens +
dist_b_mar + (1|padrao), dados)
powerPlot(Mfit)
powerPlot(Mfit, FUN = "log")
library(quantreg)
set.seed(1)
dados <- data.frame(
Area = runif(100, min = 360, max = 600)
)
dados$LVU <- 12 - .0075*dados$Area + rnorm(100, mean = 0, sd = .25)
dados$VU <- exp(dados$LVU)
medFit <- rq(VU~Area, data = dados)
powerPlot(medFit)
powerPlot(medFit, axis = "inverted")

lfpdroubi/appraiseR documentation built on April 14, 2024, 10:27 p.m.