fit.plot: Plotting fitted and empirical distribution
In barryrowlingson/opVaR: Computes operational risk

Description Usage Arguments Value See Also Examples

A function to plot empirical and fitted density (with given parameters) and compute differences between values of that densities.

1
2
3

fit.plot(x, densfun, param, distname = NULL, col = c("red"),col2 = c("grey"),col3 = c("blue"),
	 ylim = c(), xlim = c(), kernel = NULL, n = NULL, draw.diff = F, draw.max = F, 
	 scaled = F,positive=T, ...)

`x`	a data vector
`densfun`	density function
`param`	density function parameters
`distname`	density function name ( `optional`)
`col`	fitted density colour (default `red`)
`col2`	draw.diff lines colour - see draw.max option
`col3`	draw.max line colour - see draw.max option
`ylim`	`optional`
`xlim`	`optional`
`kernel`	smoothing kernel to be used; see `density`
`n`	the number of equally spaced points at which the density is to be estimated; see `density` (if not given, default `density` n is used)
`draw.diff`	logical; draw differences between empirical and estimated density values?
`draw.max`	logical; draw maximal differences between empirical and estimated density values?
`scaled`	a logical value. If TRUE, values are scaled (`x<- x/max(x)`), then values computed and after that scaled once more; try it for `dbeta` distribution
`positive`	a logival value: compute differences only for positive x density arguments?
`...`	arguments passed to `plot`

`ad`	absolute differences between empirical and fitted density values
`teor`	fitted density values
`emp`	empirical density values
`maxdiff`	maximum difference between empirical and fitted density values
`meandiff`	mean difference between empirical and fitted density values

see also loss.fit.dist, fitdistr and plot.default

# first example:

data(loss.data.object)	# data reading
x<- read.loss(1,2,loss.data.object)	# choice of data
z<- x[,2]	# z is loss data

fit.plot(z,dnorm, param = list(mean = mean(z),sd = sd(z)))

 # and the same with draw.diff = T
fit.plot(z,dnorm, param = list(mean = mean(z),sd = sd(z)),draw.diff=T)

 # draw.diff = T, draw.max = T (maximum difference is drawn), 
 # n = 40 (density values are computed only for 40 points)
fit1<-fit.plot(z,dnorm, param = list(mean = mean(z),sd = sd(z)),draw.diff=T,n=40)
length(fit1$teor) # absolute differences are computed only for positive density arguments
length(fit1$emp)  # number of arguments
# then ad is computed as sum of that absolute differences
# compare:
sum(abs(fit1$teor - fit1$emp))
fit1$ad

 # compare:
par(mfrow = c(2,1))
fit.plot(z,dnorm, param = list(mean = mean(z),sd = sd(z)),draw.diff=T,n=40,col2 = "darkblue")
fit.plot(z,dnorm, param = list(mean = mean(z),sd = sd(z)),draw.diff=T,n=40,positive=F,col2 = "darkblue")
 # more values thanks to positive = F

# second example:

 # beta distribution is special because it is scaled
fit.plot(z,dbeta,"beta", param = list(shape1 = 0.4,shape2=17)) 
 # ... and with x logarithmic scale
fit.plot(z,dbeta,"beta", param = list(shape1 = 0.4,shape2=17),log = "x") 

# third example:

parameters <-loss.fit.dist("lognormal",x)$param # fits lognormal distribution to
								# second column of our data (i.e. z)		
parameters <- as.numeric(parameters)
parameters # the estimated parameters of lognormal distribution

fit.plot(z,dlnorm,distname = "lognormal",
	param = list(meanlog = parameters[1],sdlog = parameters[2]))

 # ... and with x logarithmic scale
fit.plot(z,dlnorm,distname = "lognormal",
	param = list(meanlog = parameters[1],sdlog = parameters[2]),log="x")
 # lognormal seems to fit that data

 # and with differences drawn (there was ylim changed to see these lines better)
fit.plot(z,dlnorm,distname = "lognormal",
 param = list(meanlog = parameters[1],sdlog = parameters[2]),log="x",draw.diff=T,ylim = c(0,4e-05))