getIPMoutputDirect: Gets IPM output from a list of survival, growth, and (facultatively) fecundity objects.

Description

Get synthetic values including life expectancy, passage time, and if a fecundity matrix is available, population growth rate (lambda), stable stage disctribution, etc. This is essentially identical to getIPMoutput(),but can be a slightly swifter approach since matrices are not stored. It also allows compound matrices to be produced following levels of a discrete covariate (e.g., habitat) (using cov=TRUE). The is helpful when building multiple IPMs for study of parameter uncertainty or stochastic dynamics.

Usage

getIPMoutputDirect(survObjList, growObjList, targetSize = c(), 
	nBigMatrix, minSize, maxSize, discreteTrans = 1,cov = FALSE, 
	fecObjList = NULL, envMat = NULL, nsizeToAge = 0, 
	sizeStart = 10, integrateType = "midpoint",correction = "none", storePar=TRUE,
	chosenCov = data.frame(covariate=1), onlyLowerTriGrowth = FALSE)

Arguments

survObjList

list of survival objects.

growObjList

list of growth objects.

targetSize

Target size for passage time. If none is provided defaults to the median of the IPM meshpoints.

nBigMatrix

number of meshpoints.

minSize

minimum size.

maxSize

maximum size.

discreteTrans

object of class discreteTrans, or numeric.

cov

boolean identifying if a discrete covariate should be fitted; if cov=TRUE, an environmental matrix MUST be supplied.

fecObjList

list of fecundity objects (facultative).

envMat

environmental matrix (facultative unless cov=TRUE).

nsizeToAge

numeric describing how many size to age defined (0 - 100s).

sizeStart

starting continuous stage (e.g. size) for stage to age.

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"; option is "constant" which will multiply every column of the IPM by a constant sufficient to adjust values to those predicted for survival at that size.

storePar

boolean identifying whether you wish to store matrices of growth and survival parameters - setting "TRUE" will only work if objects are based around lms, glms, or gls

chosenCov

data-frame indicating level of the discrete covariate, or range of values where multiple covariates are modeled.

onlyLowerTriGrowth

boolean - if true, indicates that shrinking is to be avoided, i.e., any values above the diagonal should be set to zero, and the Pmatrix rescaled to continue to sum to survival. Note that this will not operate in the case of compound Pmatrices.

Value

LEmatrix of life expectancies, columns correspond to meshpoints, rows corresponding to each element of the list of P matrices
pTimematrix of passage times to the targetSize from each of the meshpoints (columns) and for each element in the P matrix list (columns).
lambdavector of population growth rates (lambda) corresponding to value obtained combining each element of the list of P matrices with the corresponding element in the list of F matrices; if no F matrix list is provided, this is a vector of NAs.
stableStagematrix of stable size distributions rows corresponding to values obtained combining each element of the list of P matrices with the corresponding element in the list of F matrices; if no F matrix list is provided, this is a matrix of NAs.
meshpointsmatrix meshpoints.
resSizematrix providing target sizes for size to age estimate (assuming age=1 at sizeStart), of length nsizeToAge space equally between the smallest and largest meshpoints.
resAgematrix providing time in time-steps to get to resSize, rows corresponding sequential elements in the list of P matrices.
surv.parmatrix of posteriors of survival parameters (columns, each sample is a row).
grow.parmatrix of posteriors of growth parameters (in columns).

References

For midpoint: Zuidema, Jongejans, Chien, During, Schieving. 2010. Integral projection models for trees: a new parameterization method and a validation of model output. Journal of Ecology 98, p345-355.

Examples

Run this code

# Example to explore variation in predictions due to parameter uncertainty, 
# using a list of IPMs generated by getListRegObjects
# Data with size and sizeNext
dff <- generateData()
gr1 <- makeGrowthObj(dff)
sv1 <- makeSurvObj(dff)

grList <- getListRegObjects(gr1, nsamp=10)
svList <- getListRegObjects(sv1, nsamp=10)


res <- getIPMoutputDirect(survObjList = svList,
                               growObjList = grList,
                               targetSize = 6,
                               nBigMatrix = 100,
                               minSize = 1.1*min(dff$size,na.rm=TRUE),
                               maxSize = 1.1*max(dff$size,na.rm=TRUE),
                               integrateType="midpoint",correction="constant", 
                               storePar=FALSE)

# Plot out with different colours for different rows:
par(mfrow = c(2,2), bty = "l", pty = "s")
plot(res$meshpoints, res$LE[1, ], xlab = "Continuous (e.g. Size) stage", 
   	ylab = "Life expectancy", type= "l", ylim = range(res$LE, na.rm=TRUE))
for (j in 1:nrow(res$LE)) points(res$meshpoints,res$LE[j, ],col=j, type = "l")

plot(res$meshpoints, res$pTime[1, ], xlab="Continuous (e.g. Size) stage", 
  	ylab = "Passage time", type = "l", ylim = range(res$pTime, na.rm=TRUE))
for (j in 1:nrow(res$pTime)) points(res$meshpoints,res$pTime[j, ],col=j,  type = "l")
     
## Only for continuous stage:
### NOT RUN ###
#dff <- generateData()

#svlist <- makePostSurvivalObjs(dataf = dff, explanatoryVariables = "size+size2", 
#	meanB = rep(0, 3), varB = rep(1e+10), nitt = 5000)
#grlist <- makePostGrowthObjs(dataf = dff, explanatoryVariables = "size+size2", 
#	meanB = rep(0, 3), varB = rep(1e+10),  nitt = 5000)
#fvlist <- makePostFecObjs(dataf=dff, explanatoryVariables="size+size2", 
#	Transform = "log", nitt = 5000)

#Just use first 10 for speed
#res <- getIPMoutputDirect(survObjList=svlist[1:10], growObjList=grlist[1:10], 
#	targetSize = 10,nBigMatrix=20, minSize = -2, maxSize = 15, 
#	cov = FALSE, fecObjList = fvlist[1:10],
#	envMat = NULL, nsizeToAge = 5, sizeStart = -1)

#names(res)

# Plot out with different colours for different rows:
#par(mfrow = c(2,2), bty = "l", pty = "s")
#plot(res$meshpoints, res$LE[1, ], xlab = "Continuous (e.g. Size) stage", 
#	ylab = "Life expectancy", type= "l", ylim = range(res$LE, na.rm=TRUE))
#for (j in 1:nrow(res$LE)) points(res$meshpoints,res$LE[j, ],col=j, type = "l")

#plot(res$meshpoints, res$pTime[1, ], xlab="Continuous (e.g. Size) stage", 
#	ylab = "Passage time", type = "l", ylim = range(res$pTime, na.rm=TRUE))
#for (j in 1:nrow(res$pTime)) points(res$meshpoints,res$pTime[j, ],col=j,  type = "l")

#plot(res$meshpoints,Re(res$stableStage[1, ]), xlab = "Size", 
#	ylab = "Stable stage distribution", type = "l", 
#		ylim = range(Re(res$stableStage), na.rm = TRUE))
#for (j in 1:nrow(res$stableStage)) points(res$meshpoints, 
#	Re(res$stableStage[j, ]), col = j, type = "l")

#hist(res$lambda, xlab = expression(lambda), ylab = "", main = "", col = "grey")

## For continuous (e.g. size) and discrete (e.g. seedbank) stages:
#dff <- generateDataDiscrete()

#svlist <- makePostSurvivalObjs(dataf=dff, explanatoryVariables="size+size2", 
#	meanB = rep(0, 3), varB = rep(1e+10), nitt = 5000)
#grlist <- makePostGrowthObjs(dataf=dff, explanatoryVariables = "size+size2", 
#	meanB = rep(0, 3), varB = rep(1e+10), nitt = 5000)
#fvlist <- makePostFecObjs(dataf = dff, explanatoryVariables = "size+size2", 
#	Transform = "log", nitt = 5000)

#res <- getIPMoutputDirect(survObjList = svlist[1:10], growObjList = grlist[1:10], 
#	targetSize = 10, nBigMatrix = 20, minSize = -2, maxSize = 15, 
#      discreteTrans=makeDiscreteTrans(dff), cov = FALSE, 
#		fecObjList = fvlist[1:10], nsizeToAge = 5, sizeStart = -1)