Skip to content

Fitted Value Pareto Distribution

4 messages · livia, Markus Jäntti, J. R. M. Hosking

#
I would like to fit a Pareto Distribution and I am using the following codes. 

I thought the fitted (fit1) should be the fitted value for the data, is it
correct? As the result of the "fitted" turns out to be a single value for
all. 

fit=vglm(ycf1 ~ 1, pareto1(location=alpha), trace=TRUE, crit="c") 
fitted(fit) 

The result is 
fitted(fit)
            [,1]
 [1,] 0.07752694
 [2,] 0.07752694
 [3,] 0.07752694
 [4,] 0.07752694
 [5,] 0.07752694
 [6,] 0.07752694
 [7,] 0.07752694
 [8,] 0.07752694
 [9,] 0.07752694
[10,] 0.07752694
[11,] 0.07752694
[12,] 0.07752694
[13,] 0.07752694

Could anybody give me some advice?
#
livia wrote:
Your model only includes an intercept, so the fitted value  is supposed to be 
the same for all units (there is nothing in your model that allows the fitted 
value to vary across units).

markus
#
livia wrote:
I don't have whatever package function 'vglm' comes from (did you
follow the instructions in the last two lines of your post?), but you
can fit a GPD and get fitted values for it by some such approach as
this:

   library(POT)
   threshold <- 0  # probably
   para <- fitgpd(ycf1, threshold, method="pwmu")$param
   ycf1.fit <- qgpd( ppoints(ycf1, a=0.44), threshold, para[1], para[2])

Note that the above code contains my own preferences for fitting
method and plotting positions: yours may differ.


J. R. M. Hosking
#
Thank you very much and that is exactly what I am looking for. Another
question would be how can I test the goodness of fit for the Pareto
distribution?
J. Hosking wrote: