Hi all,
I've tried to plot a vector which has two peaks in the density.
This link shows the figure.
http://docs.google.com/View?docid=dcvdrfrh_1dk9r2rc7
The red line is normal curve and green line is gamma curve.
Notice that red line can correctly fit the histogram that has two peaks
(i.e. red curve also has two peaks).
But the gamma curve there only has one curve.
Is there a way I can fit the gamma function such that it also
yields two peaks?
Below is my code that generate the graph:
__BEGIN__
dgexp <- density(gexp.arr)
gexp.ymax <- max(dgexp$y)
# Histograms
gehist <- hist(gexp.arr,
col="blue",
main=paste(genm,"-",desc, ",
Mean=",ge_mean, ", SD=",ge_sd, ",
Krt=",ge_kurt, ", AIC=",aic, ",
BIC=",bic, sep=""),
more = TRUE,
xlab = "Exp Level",
ylim = c(0, gexp.ymax),
freq=FALSE
)
# Normal Curve
lines(density(gexp.arr), col = "red",lwd = 3)
# Gamma Curve
curve(dgamma(x,1.25,gexp.ymax),add=TRUE,col="green",lwd=3)
__END__
- Gundala Viswanath
Jakarta - Indonesia
Gundala Viswanath <gundalav <at> gmail.com> writes:> > Hi all, > > I've tried to plot a vector which has two peaks in the density. > This link shows the figure. > > http://docs.google.com/View?docid=dcvdrfrh_1dk9r2rc7 > > The red line is normal curve and green line is gamma curve. > Notice that red line can correctly fit the histogram that has two peaks > (i.e. red curve also has two peaks). > > But the gamma curve there only has one curve. > > Is there a way I can fit the gamma function such that it also > yields two peaks? >The gamma distribution is always unimodal (i.e., a single peak). (You call the red line a "normal curve", but it's really a nonparametric density estimate.) You might want to search the list and the available packages for "mixture models", which are a standard way of fitting multimodal distributions ... Ben Bolker