Created
June 26, 2012 02:55
-
-
Save emhart/2992981 to your computer and use it in GitHub Desktop.
Code from my curve fitting blog post about Ecolog.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
###### The exponential growth of ecolog | |
###### Raw data can be found here | |
###### https://listserv.umd.edu/cgi-bin/wa?A2=ind1108e&L=ecolog-l&P=3401 | |
#####Data from Aug 31. 2011 post | |
require(date) | |
subscribers <- c(100,6000,7000,8000,9000,10000,11000,12000,13000) | |
start.date <- mdy.date(1,1,92) | |
month <- c("01/01/1992","09/01/2006","11/01/2007","10/01/2008","03/01/2009","04/01/2010","09/01/2010","02/01/2011","09/01/2011") | |
days <- as.date(month)-start.date | |
z <- nls(subscribers~100*exp(r*days),start=list(r=.001)) | |
######Plot the data #### | |
plot(days,subscribers,xlab="Days since Sept 1992",ylab="Number of Subscribers",xlim=c(0,7500),ylim=c(0,18000)) | |
days.v <- 1:8000 | |
fit.z <- 100*exp(coef(z)*days.v) | |
lines(days.v,fit.z) | |
#### now add the new data point #### | |
new.day <- mdy.date(6,23,2012)-start.date | |
points(days.v[new.day],fit.z[new.day],col=2,pch=19) | |
points(new.day,15000,pch=19,col=1) | |
#### so how can we find an estimation of what the starting value should be? | |
dat <- cbind(days[-1],subscribers[-1]) | |
fp.est <- function(param,dat){ | |
z <- nls(dat[,2]~param*exp(r*dat[,1]),start=list(r=.001)) | |
return(deviance(z)) | |
} | |
tmp <- optim(1,fp.est,dat=dat,method="L-BFGS-B",lower=1,upper=6000) | |
##### Ok, can we improve with a jacknife style estimate? ###### | |
##### Generate all the possibe combinations | |
comb.mat <- combn(1:9,8) | |
dat <- rbind(dat,c(new.day,15000)) | |
est <- vector() | |
for(i in 1:dim(comb.mat)[2]){ | |
my.dat <- dat[comb.mat[,i],] | |
tmp <- optim(1,fp.est,dat=my.dat,method="L-BFGS-B",lower=1,upper=6000) | |
est[i] <- tmp$par | |
} | |
tmp <- optim(1,fp.est,dat=dat,method="L-BFGS-B",lower=1,upper=6000) | |
####Now calculate the bias and variance from the jackknife | |
bias <- -(1/8)*sum(8*(tmp$par-est)) | |
var_jack <- (1/(8*9))*sum((8*(tmp$par-est))^2-(9*bias)^2) | |
sd_jack <- sqrt(var_jack) | |
#### Make the final plot | |
z <- nls(dat[,2]~round(tmp$par)*exp(r*dat[,1]),start=list(r=.001)) | |
######Plot the data #### | |
plot(dat,xlab="Days since Sept 1992",ylab="Number of Subscribers",xlim=c(0,7500),ylim=c(0,18000)) | |
days.v <- 1:8000 | |
fit.z <- round(tmp$par)*exp(coef(z)*days.v) | |
lines(days.v,fit.z) | |
#### now add the new data point #### | |
new.day <- mdy.date(6,23,2012)-start.date | |
points(days.v[new.day],fit.z[new.day],col=1,pch=19) | |
points(new.day,15000,pch=19,col=2) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment