graphresU.fun: Validation analysis of PP uniform (generalized) residuals

Description

This function checks the properties that must be fulfilled by the uniform (generalized) residuals of a PP: uniform character and uncorrelation. Optionally, the existence of patterns versus covariates or potentially influent variables can be graphically analyzed.

Usage

graphresU.fun(unires, posE,  Xvariables = NULL, namXv = NULL, flow = 0.5,
 tit = "", addlow = TRUE, histWgraph=TRUE, plotDisp=c(2,2), indgraph = FALSE)

Arguments

unires: Numeric vector of the uniform residuals.
posE: Numeric vector of the occurrence times of the PP.
Xvariables: Matrix of variables to perform the residual plots (each column is a variable).
namXv: Optional. Vector of names of the variables in Xvariables.
tit: Character string. A title for the plot.
addlow: Logical flag. If it is TRUE, a lowess is added to the plots.
flow: Argument f for the lowess smoother; see lowess for details.
histWgraph: Logical flag. If it is TRUE, a new graphical device is opened with the option record=TRUE, so that the history of all plots is recorded in the new device. This option may not work on some platforms; for example, RStudio does not allow the user to open new graphical devices.
plotDisp: A vector of the form c(nr, nc). The residual versus variables plots will be drawn in a nr\(\times\)nc array. It is used as argument mfrow in par. By default, a 2 \(\times\) 2 layout is used.
indgraph: Logical flag. If it is TRUE, the validation plots (except the residuals versus variables plots) are carried out in four1 \(\times\) 1 layouts. By default, a 2 \(\times\) 2 layout is used.

Details

The validation analysis of the uniform character consists in a uniform Kolmogorov-Smirnov test and a qqplot with a 95% confidence band based on a beta distribution. The analysis of the serial correlation is based on the Pearson correlation coefficient, Ljung-Box tests and a lagged serial correlation plot. An index plot of the residuals and residual plots versus the variables in argument Xvariables are performed to analyze the effect of covariates or other potentially influent variables. These plots will show if the mean or dispersion of the residuals vary sistematically, see model diagnostic of Cox-Snell residuals in Collett (1994) for more details.

References

Abaurrea, J., Asin, J., Cebrian, A.C. and Centelles, A. (2007). Modeling and forecasting extreme heat events in the central Ebro valley, a continental-Mediterranean area. Global and Planetary Change, 57(1-2), 43-58.

Baddeley, A., Turner, R., Moller, J. and Hazelton, M. (2005). Residual analysis for spatial point processes. Journal of the Royal Statistical Society, Series B, 67, 617-666.

Cebrian, A.C., Abaurrea, J. and Asin, J. (2015). NHPoisson: An R Package for Fitting and Validating Nonhomogeneous Poisson Processes. Journal of Statistical Software, 64(6), 1-24.

Collett, D. (1994). Modelling survival data in medical research. Chapman and Hall.

Ogata, Y. (1988). Statistical models for earthquake occurrences and residual analysis for point processes. Journal of the American Statistical Association, 83(401), 9-27.

Examples

Run this code


#Since  only one graphical device is opened  and the argument histWgraph 
#is TRUE by default, the resulting residual plots  (three pages with the 
#considered 1X2 layout for the residual versus variables plot)   
#can be scrolled up and down with the "Page Up" and "Page Down" keys.

X1<-rnorm(500)
X2<-rnorm(500)

graphresU.fun(unires=runif(30,0,1),posE=round(runif(30,0,500)), 
  Xvariables=cbind(X1,X2), namXv=c("X1","X2"),tit="Example",flow=0.7,plotDisp=c(1,2))