Goodness of fit for gamma distributions
It sounds like you just want to graph it though. For gammas, it's nice to graph the log of the density, because the tail is so thin and long, so you don't see much otherwise: mydata <- rgamma(10000, shape=1.1, rate=2.5) # now suppose you fit a gamma distribution, and get these estimated parameters: shapeest <- 1.101 rateest <- 2.49 h <- hist(mydata, breaks=50, plot=FALSE) plot(h$mids, log(h$density)) curve(log(dgamma(x, shape=shapeest, rate=rateest)), add=TRUE) #Remko ------------------------------------------------- Remko Duursma Post-Doctoral Fellow Centre for Plant and Food Science University of Western Sydney Hawkesbury Campus Richmond NSW 2753 Dept of Biological Science Macquarie University North Ryde NSW 2109 Australia Mobile: +61 (0)422 096908
On Wed, Jan 28, 2009 at 1:13 AM, Dan31415 <d.m.mitchell at reading.ac.uk> wrote:
Thanks for that Remko, but im slightly confused because isnt this testing the goodness of fit of 2 slightly different gamma distributions, not of how well a gamma distribution is representing the data. e.g. data.vec<-as.vector(data) (do some mle to find the parameters of a gamma distribution for data.vec) xrarea<-seq(-2,9,0.05) yrarea<-dgamma(xrarea,shape=7.9862,rate=2.6621) so now yrarea is the gamma distribution and i want to compare it with data.vec to see how well it fits. regards, Dann Remko Duursma-2 wrote:
Hi Dann, there is probably a better way to do this, but this works anyway: # your data gamdat <- rgamma(10000, shape=1, rate=0.5) # comparison to gamma: gamsam <- rgamma(10000, shape=1, rate=0.6) qqplot(gamsam,gamdat) abline(0,1) greetings Remko ------------------------------------------------- Remko Duursma Post-Doctoral Fellow Centre for Plant and Food Science University of Western Sydney Hawkesbury Campus Richmond NSW 2753 Dept of Biological Science Macquarie University North Ryde NSW 2109 Australia Mobile: +61 (0)422 096908 On Tue, Jan 27, 2009 at 3:38 AM, Dan31415 <d.m.mitchell at reading.ac.uk> wrote:
I'm looking for goodness of fit tests for gamma distributions with large data sizes. I have a matrix with around 10,000 data values in it and i have fitted a gamma distribution over a histogram of the data. The problem is testing how well that distribution fits. Chi-squared seems to be used more for discrete distributions and kolmogorov-smirnov seems that large sample sizes make it had to evaluate the D statistic. Also i haven't found a qq plot for gamma, although i think this might be an appropriate test. in summary -is there a gamma goodness of fit test that doesnt depend on the sample size? -is there a way of using qqplot for gamma distributions, if so how would you calculate it from a matrix of data values? regards, Dann -- View this message in context: http://www.nabble.com/Goodness-of-fit-for-gamma-distributions-tp21668711p21668711.html Sent from the R help mailing list archive at Nabble.com.
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
-- View this message in context: http://www.nabble.com/Goodness-of-fit-for-gamma-distributions-tp21668711p21686095.html Sent from the R help mailing list archive at Nabble.com.
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.