ddcomp: Evaluate the fit of a tree diameter distribution

In lmfor: Functions for Forest Biometrics

Evaluate the fit of a tree diameter distribution

Description

A function to compare the fit of the observed tree diameter data (d) to a specified diameter distribution (density).

Usage

ddcomp(d,density="dweibull",power=0,limits=seq(0,100),limitsd=limits,plot=FALSE,...)

Arguments

d	numeric vector of observed diameters
density	either a valid name for a probability density function in R or a vector of diameter class densities for diameter classes whose limits are given in vector limitsd
power	the weight used in error index. Value 2 gives BA weight, 0 (default) the unweighted
limits	the diameter class limits to compute the error index
limitsd	see the description of argument density
plot	logical. Should a graph be produced to illustrate the ecdf of d and the cdf corresponding to density
...	additional arguments passed to function specified by a character-type density. e.g. Weibull shape and scale of if density="dweibull"

Details

The comparison is done for mean, variance and standard deviation and shape. The shape is compared by computing the sum of absolute differences (error index) in densities for the observed data and predicted density in diameter classes specified by "limits". The error index has therefore a value between 0 (complete match) and 2 (complete mismatch). The error index is computed for the predicted density as such (ei1) and to a rescaled and switched density, which has exactly same mean and variance as the given diameter data (e12).

The error index is calculated as the sum of variable (f_{obs}-f_{pred})x^{power} over the diameter classes, where x is the midpoint of the diaemeter class and f_{obs}-f_{pred} is the difference in predicted and observed frequency. By default, power=0.

Value

A list of components

mudif	The difference in means
vardif	The difference in variances
sddif	The difference in standard deviations
ei1	the error index for original predicted distribution (see details)
ei1	the error index for scaled predicted distribution (see details)

Author(s)

Lauri Mehtatalo <[e-mail]>

Examples

# Example 
# Observed diameters 
d<-c(18.8,24.2,18.7,13.0,18.9,22.4,17.6,22.0,18.8,22.9,
     16.7,13.7,20.6,15.1,31.8,17.2,19.6,16.8,19.3,27.4,
	 23.7,18.2,19.7,18.9,23.0,21.4,23.8,22.1,24.2,20.9)

# Weibull(5,20) distribution in 1 cm classes (class limits from 0,...,60)
f<-pweibull(1:60,5,20)-pweibull(0:59,5,20)

# compare using the classified true distribution (approximate)
ddcomp(d,density=f,limitsd=0:60,limits=0:100,plot=TRUE)

# compare b specifying a Weibull dsitribution (accurate)
ddcomp(d,density="dweibull",shape=5,scale=20,plot=TRUE)

lmfor documentation built on April 30, 2022, 1:08 a.m.