IPMpack (version 2.1)

sensParams: Estimates sensitivity and elasticity of lambda (or R0, or Life expectancy of a chosen bin) to parameters underlying an IPM.

Description

Uses perturbation to estimate the sensitivity and elasticity of all the parameters underlying an IPM.

Usage

sensParams(growObj, survObj, fecObj=NULL, clonalObj=NULL, 
	nBigMatrix, minSize, maxSize,  
	chosenCov = data.frame(covariate = 1), discreteTrans = 1, 
	integrateType = "midpoint", correction = "none", preCensusFec = TRUE, 	
	postCensusSurvObjFec = NULL, postCensusGrowObjFec = NULL,  
	preCensusClonal = TRUE, postCensusSurvObjClonal = NULL, 
	postCensusGrowObjClonal = NULL, delta = 1e-04, 
	response="lambda", chosenBin=1)

Arguments

growObj

a growth object.

survObj

a survival object.

fecObj

a fecundity object (not necessary for life expectancy analysis).

clonalObj

a clonality object (not necessary for life expectancy analysis).

nBigMatrix

numeric, number of bins of size used in the IPM matrix.

minSize

numeric, minimum size used for meshpoints of the IPM matrix.

maxSize

numeric, maximum size used for meshpoints of the IPM matrix.

chosenCov

level or value of the covariate(s) at which sensitivity estimation is desired

discreteTrans

matrix of discrete transitions; or 1 if there is none

integrateType

integration type, defaults to "midpoint" (which uses probability density function); other option is "cumul" (which uses the cumulative density function)

correction

correction type, defaults to none. The first option is constant which will multiply every column of the IPM by a constant sufficient to adjust values to those predicted for total fertility at that size. The second option is discretizeExtremes which will place all transitions to sizes smaller than minSize into the smallest bin, and transitions to sizes larger than maxSize into the largest bin.

preCensusFec

logical (TRUE or FALSE), indicating whether the fecundity object represents an interval between pre-breeding or a post-breeding censusses. Defaults to TRUE (pre-breeding census), meaning that all reproduction and offspring rates required for the F matrix are embedded in fecObj. Alternatively, an F matrix based on post-breeding census (preCensusFec=FALSE) uses postCensusSurvObjFec and postCensusGrowObjFec, to cover the survival and growth of the parents until the reproduction event. (not necessary for life expectancy analysis)

postCensusSurvObjFec

survival object representing the survival of the parents until the reproduction event. If not specified (and preCensusFec = FALSE) it is assumed that all parents survive until the reproduction event. (not necessary for life expectancy analysis)

postCensusGrowObjFec

growth object representing the growth of surviving parents until the reproduction event. If not specified (and preCensusFec = FALSE) it is assumed that the parents do not grow until the reproduction event. (not necessary for life expectancy analysis)

preCensusClonal

logical (TRUE or FALSE), indicating whether the clonality object represents an interval between pre-breeding or a post-'breeding' censusses. Defaults to TRUE (pre-'breeding' census), meaning that all clonal propagation and offspring rates required for the C matrix are embedded in clonalObj. Alternatively, an C matrix based on post-'breeding' census (preCensusClonal=FALSE) uses postCensusSurvObjClonal and postCensusGrowObjClonal, to cover the survival and growth of the parents until the clonal propagation event. (not necessary for life expectancy analysis)

postCensusSurvObjClonal

survival object representing the survival of the parents until the clonal propagation event. If not specified (and preCensusClonal = FALSE) it is assumed that all parents survive until the clonal propagation event. (not necessary for life expectancy analysis)

postCensusGrowObjClonal

growth object representing the growth of surviving parents until the clonal propagation event. If not specified (and preCensusClonal = FALSE) it is assumed that the parents do not grow until the clonal propagation event. (not necessary for life expectancy analysis)

delta

size of the perturbation desired

response

whether lambda, R0 or life expectancy of a desired bin (lifeExpect with chosenBin) is required

chosenBin

for analysis of life expectancy, which bin in the IPM Life expectancy should be compared for

Value

sens

a vector of sensitivities of lambda or other variable with names corresponding to parameters.

elas

a vector of elasticities to lambda or other variable with names corresponding to parameters.

Details

The values returned by sensParam are calculated by first calculating lambda for the chosen IPM; then modifying the focal parameter c by a very small amount, c.new=c*(1+delta) (the default for delta =1e-4, but users may specify the value that they want). The function then rebuilds the T and F matrices, and re-calculates lambda. Sensitivity is calculated as: sens = df(x)/dx = (lam.new-lam)/(c*delta) i.e., the function estimates the degree to which a small change in the parameter results in a small change in lambda; and elasticity is calculated as: elas = sens*c/lam = (lam.new-lam)/(lam*delta) which corresponds to the proportional change in lambda as an outcome of the proportional change in the parameter; analagous calculations are used for R0 and life expectancy.

NOTE: in previous versions of IPMpack (pre 2.0), the output of this function was mis-aligned.

References

Rees and Rose. 2002. Evolution of flowering strategies in Oenothera glazioviana: an integral projection model approach. Proceedings of the Royal Society London Seres B 269, p1509-1515.

See Also

sens, elas

Examples

dff <- generateData()

#lambda
res <- sensParams(growObj = makeGrowthObj(dff), 
survObj = makeSurvObj(dff), fecObj = makeFecObj(dff, Transform="log"), 
nBigMatrix = 50, minSize = min(dff$size, na.rm=TRUE), 
maxSize = max(dff$size, na.rm = TRUE))

par(mfrow = c(2, 1), bty = "l", pty = "m") 
barplot(res$sens, 
main = expression("Parameter sensitivity of population growth rate "* lambda), 
las = 2, cex.names = 0.5) 
barplot(res$elas, 
main = expression("Parameter elasticity of population growth rate "* lambda), 
las = 2, cex.names = 0.5) 

#R0
resR0 <- sensParams(growObj = makeGrowthObj(dff), 
survObj = makeSurvObj(dff), fecObj = makeFecObj(dff, Transform="log"), 
nBigMatrix = 50, minSize = min(dff$size, na.rm=TRUE), 
maxSize = max(dff$size, na.rm = TRUE), response="R0")

par(mfrow = c(2, 1), bty = "l", pty = "m") 
barplot(resR0$sens, 
main = expression("Parameter sensitivity of net reproductive rate R"[0]), 
las = 2, cex.names = 0.5) 
barplot(resR0$elas, 
main = expression("Parameter elasticity of net reproductive rate R"[0]), 
las = 2, cex.names = 0.5) 

#life expectancy
resLE <- sensParams(growObj = makeGrowthObj(dff), 
survObj = makeSurvObj(dff),  nBigMatrix = 50, 
minSize = min(dff$size, na.rm=TRUE), maxSize = max(dff$size, na.rm =
TRUE), chosenBin=1, response="lifeExpect")

par(mfrow = c(2, 1), bty = "l", pty = "m") 
barplot(resLE$sens, 
main = expression("Parameter sensitivity of Life Expectancy"*eta[0]), 
las = 2, cex.names = 0.5) 
barplot(resLE$elas, 
main = expression("Parameter elasticity of Life expectancy"*eta[0]), 
las = 2, cex.names = 0.5) 

# Same as lambda above, but with two fecundity functions
dff$fec2 <- dff$fec>0 #create binomial describing e.g., prob of flowering
dff$fec[dff$fec==0] <- NA #take out zeros to avoid Inf when fit with log
fv1 <- makeFecObj(dff, Formula = c(fec~size+size2,fec2~size), 
    Transform=c("log","none"),Family = c("gaussian","binomial"))

res <- sensParams(growObj=makeGrowthObj(dff), survObj = makeSurvObj(dff), 
fecObj = fv1, nBigMatrix = 50, minSize = min(dff$size, na.rm = TRUE), 
maxSize = max(dff$size, na.rm = TRUE))

par(mfrow = c(2, 1), bty = "l", pty = "m") 
barplot(res$sens, 
main = expression("Parameter sensitivity of population growth rate " *lambda), 
las = 2, cex.names = 0.5) 
barplot(res$elas, 
main = expression("Parameter elasticity of population growth rate " *lambda), 
las = 2, cex.names = 0.5) 

# Same but with two fecundity functions and a constant
fv1@fecConstants[1] <-0.5
res <- sensParams(growObj = makeGrowthObj(dff), survObj = makeSurvObj(dff), 
fecObj = fv1, nBigMatrix = 50, minSize = min(dff$size, na.rm = TRUE), 
maxSize = max(dff$size, na.rm = TRUE))

par(mfrow = c(2, 1), bty = "l", pty = "m") 
barplot(res$sens, 
main = expression("Parameter sensitivity of population growth rate " *lambda), 
las = 2, cex.names = 0.5) 
barplot(res$elas, 
main = expression("Parameter elasticity of population growth rate " *lambda), 
las = 2, cex.names = 0.5)

# Same but with a discrete class
dff <- generateData(type="discrete")
res <- sensParams(growObj = makeGrowthObj(dff), survObj = makeSurvObj(dff), 
	fecObj = makeFecObj(dff), discreteTrans=makeDiscreteTrans(dff),
	nBigMatrix = 50, minSize = min(dff$size, na.rm = TRUE), 
	maxSize = max(dff$size, na.rm = TRUE))

par(mfrow = c(2, 1), bty = "l", pty = "m") 
barplot(res$sens, 
main = expression("Parameter sensitivity of population growth rate " *lambda), 
las = 2, cex.names = 0.5) 
barplot(res$elas, 
main = expression("Parameter elasticity of population growth rate " *lambda), 
las = 2, cex.names = 0.5)