phenology (version 7.3)

fit_phenology: Fit the phenology parameters to timeseries of counts.

Description

Function of the package phenology to fit parameters to timeseries. To fit data, the syntaxe is : Result <- fit_phenology(data=dataset, fitted.parameters=par, fixed.parameters=pfixed, trace=1, method_incertitude=2, zero_counts=TRUE, hessian=TRUE) or if no parameter is fixed : Result <- fit_phenology(data=dataset, fitted.parameters=par) Add trace=1 [default] to have information on the fit progression or trace=0 to hide information on the fit progression. zero_counts = c(TRUE, TRUE, FALSE) indicates whether the zeros have been recorded for each of these timeseries. Defaut is TRUE for all. hessian = FALSE does not estimate se of parameters. If the parameter Theta is fixed to +Inf, a Poissonian model of daily nest distribution is implemented.

Usage

fit_phenology(data = file.choose(), fitted.parameters = NULL,
  fixed.parameters = NULL, method_incertitude = "convolution",
  infinite = 200, zero_counts = TRUE, store.intermediate = FALSE,
  file.intermediate = "Intermediate.rda", parallel = TRUE,
  hessian = TRUE, silent = FALSE, growlnotify = TRUE,
  cofactors = NULL, add.cofactors = NULL, zero = 1e-09, lower = 0,
  upper = Inf, control = list(trace = 1, REPORT = 1, maxit = 1000),
  method = c("Nelder-Mead", "L-BFGS-B"))

Arguments

data

A dataset generated by add_format

fitted.parameters

Set of parameters to be fitted

fixed.parameters

Set of fixed parameters

method_incertitude

'combinatory' estimates likelihood of all combinations for nest numbers; 'convolution' [default] uses the exact likelihood of the sum of negative binomial distribution.

infinite

Number of iterations for dSnbinom() used for method_incertitude='convolution'

zero_counts

example c(TRUE, TRUE, FALSE) indicates whether the zeros have been recorded for each of these timeseries. Defaut is TRUE for all.

store.intermediate

TRUE or FALSE to save the intermediates

file.intermediate

Name of the file where to save the intermediates as a list

parallel

If FALSE, no parallel computing is done to evaluate likelihood

hessian

If FALSE does not estimate se of parameters

silent

If TRUE does not show any message

growlnotify

If FALSE, does not send growl notification (only in MacOSX)

cofactors

data.frame with a column Date and a column for each cofactor

add.cofactors

Names of the column of parameter cofactors to use as a cofactor

zero

If the theoretical nest number is under this value, this value wll be used

lower

Lower bound for each parameter

upper

Upper bound for each parameter

control

List for control parameters for optim

method

Method used by optim. Several can be setup.

Value

Return a list of with data and result

Details

fit_phenology fits parameters to timeseries.

Examples

# NOT RUN {
library(phenology)
# Read a file with data
Gratiot <- read.delim("http://max2.ese.u-psud.fr/epc/conservation/BI/Complete.txt", 
header=FALSE)
data(Gratiot)
# Generate a formatted list nammed data_Gratiot 
data_Gratiot <- add_phenology(Gratiot, name="Complete", 
		reference=as.Date("2001-01-01"), format="%d/%m/%Y")
# Generate initial points for the optimisation
parg <- par_init(data_Gratiot, fixed.parameters=NULL)
# Run the optimisation
result_Gratiot <- fit_phenology(data=data_Gratiot, 
		fitted.parameters=parg, fixed.parameters=NULL)
data(result_Gratiot)
# Plot the phenology and get some stats
output <- plot(result_Gratiot)
# Use fit with co-factor
# First extract tide information for that place
td <- tide.info(year=2001, latitude=4.9167, longitude=-52.3333, tz="America/Cayenne")
td2 <- td[td$Tide=="High Tide", ]
td3 <- cbind(td2, Date=as.Date(td2$Date.Time))
td4 <- td3[(as.POSIXlt(td3$Date.Time)$hou<6) | (as.POSIXlt(td3$Date.Time)$hou>18), ]
with(td4, plot(Date.Time, Level, type="l"))
data_Gratiot$Complete$Date
td5 <- merge(data_Gratiot$Complete, td4, by.x="Date", by.y="Date")
td6 <- td5[, c("Date", "Level")]
parg <- par_init(data_Gratiot, fixed.parameters=NULL, add.cofactors="Level")
result_Gratiot_CF <- fit_phenology(data=data_Gratiot, 
		fitted.parameters=parg, fixed.parameters=NULL, cofactors=td6, 
		add.cofactors="Level")
compare_AIC(WithoutCF=result_Gratiot, WithCF=result_Gratiot_CF)
plot(result_Gratiot_CF)

# Example with two series fitted with different peaks but same Length of season

Gratiot2 <- Gratiot
Gratiot2[, 2] <- floor(Gratiot2[, 2]*runif(n=nrow(Gratiot2)))
data_Gratiot <- add_phenology(Gratiot, name="Complete",
                              reference=as.Date("2001-01-01"), format="%d/%m/%Y")
data_Gratiot <- add_phenology(Gratiot2, name="Complete2",
                              reference=as.Date("2001-01-01"), 
                              format="%d/%m/%Y", previous=data_Gratiot)
pfixed=c(Min=0)
p <- par_init(data_Gratiot, fixed.parameters = pfixed)
p <- c(p, Peak_Complete=175, Peak_Complete2=175)
p <- p[-4]
p <- c(p, Length=90)
p <- p[-(3:4)]
result_Gratiot <- fit_phenology(data=data_Gratiot, fitted.parameters=p, 
fixed.parameters=pfixed)
# An example with bimodality
g <- Gratiot
g[30:60, 2] <- sample(10:20, 31, replace = TRUE)
data_g <- add_phenology(g, name="Complete", reference=as.Date("2001-01-01"), 
                        format="%d/%m/%Y")
parg <- c('Max.1_Complete' = 5.6344636692341856, 
          'MinB.1_Complete' = 0.15488810581002324, 
          'MinE.1_Complete' = 0.2, 
          'LengthB.1' = 22.366647176407636, 
          'Peak.1' = 47.902473939250036, 
          'LengthE.1' = 17.828495918533015, 
          'Max.2_Complete' = 33.053364083447434, 
          'MinE.2_Complete' = 0.42438173496989717, 
          'LengthB.2' = 96.651564706802702, 
          'Peak.2' = 175.3451874571835, 
          'LengthE.2' = 62.481968743789835, 
          'Theta' = 3.6423908093342572)
pfixed <- c('MinB.2_Complete' = 0, 
          'Flat.1' = 0, 
          'Flat.2' = 0)
result_g <- fit_phenology(data=data_g, fitted.parameters=p0, fixed.parameters=pfixed)
plot(result_g)
# }