Dear All,
(this msg is a statistics/computing question to the list)
I'm trying to implement a modern-version of a (classic) "several-step
protocol" in Fishery Biology (due to Bhattacharya, 1967): analysis of
length-frequency distribution of fish larvae to id cohorts and later estimate
growh rates!
I've trouble with the 1st step: using kernel density estimation (KDE)
function to examine size-freq distributions (variable sl):
> density(sl[dofyr==139],bw="nrd0") (dofyr means day of year and
refers to a particular sampling date)
> kd139
Call:
density.default(x = sl[dofyr == 139], bw = "nrd0")
Data: sl[dofyr == 139] (55 obs.); Bandwidth 'bw' = 0.4234
x y
Min. : 4.870 Min. :0.0001924
1st Qu.: 7.207 1st Qu.:0.0120950
Median : 9.545 Median :0.0406854
Mean : 9.545 Mean :0.1068383
3rd Qu.:11.883 3rd Qu.:0.1973223
Max. :14.220 Max. :0.3491222
Now the question: the densities per x value (obtained using kd139$y) do not sum
to one? I expected them to be probabilities and wanted to "extract"
the smoothed frequencies for the midpoints of size-intervals (in this example
each class is 0.4234 mm wide?!)
Thanks in advance,
Eduardo Esteves
University of Algarve School of Technology (Portugal)
ps-I'm trying implement the other steps of the procedure, so it's
possible that I'll need help (again)!
[[alternative HTML version deleted]]
