grasp {grasp}R Documentation

Generalized Regression Analysis and Spatial Prediction

Description

GRASP is a general method for making spatial predictions of response variables (RV) using point surveys of the RV and spatial coverages of Predictor variables (PV). GRASP uses Trevor Hastie's gam library.

Usage

grasp(selected.responses, selected.predictors, title = "GRASP: ", path = "", save.outputs = FALSE, save.PNG = FALSE, 
gr.fam = "binomial", weights = FALSE, set.limits = FALSE, use.limits = FALSE, reset.lim = TRUE, selxlim = gr.selX, npast = 10,
make.summary = FALSE, plot.maps = FALSE, plot.distry = FALSE, plot.histograms = FALSE, plot.respvspred = FALSE, plot.xpred = FALSE,
plot.correlation = FALSE, corlim = 100, use.correlation = FALSE, reset.cor = TRUE, full.models = FALSE, stepwise.models = FALSE, 
test = "F", plimit = 0.05, direction = "both", df1 = 0, df2 = 4, startwith = gr.selX, resetstart = TRUE,
contributions = FALSE, plot.contributions = FALSE, plot.models = FALSE, response.curves = FALSE, model.anova = FALSE, validate.models = FALSE, 
predictions = FALSE, export.predictions = FALSE, plot.predictions = FALSE, lookup.tables = FALSE, 
plotpar = c(3, 3), prescol= "black", abscol="grey", nbar = 10, show.details = FALSE, StdError = TRUE, cvgroups = 5, resolution = as.integer((max(XXXpred$CX) - min(XXXpred$CX)) / (length(unique(XXXpred$CX)) - 1)) )

Arguments

selected.responses Vector of column numbers or column names of YYY table, which contains response variables
selected.predictors Vector of column numbers or column names of XXX table, which contains predictor variables
title Title used in GRASP outputs
path Working directory path
save.outputs Send outputs for each response to a specific log file in your working directory
save.PNG Save graphs as PNG files in your working directory
gr.fam Choice of statistical distribution for Ys: "binomial", "poisson", "quasibinomial", "quasipoisson", "gaussian"
weights Recalculate weights of 0s so that their sum is equal to sum of 1s
set.limits Calculate and save limits of the environmental envelop of each responses (Ys)
use.limits Use the saved limits in anylyses
reset.lim Reset saved limits
selxlim Vector of column numbers or column names of XXX table used to define limits
npast Number of absences (0s) kept after the last presence (1) along each selected predictors used for calculating limits
make.summary Calculate the data summary of selected Ys and Xs
plot.maps Plot the spatial distribution of Ys
plot.distry Histograms of the distribution of Ys
plot.histograms Plot the distribution of Ys on histograms of Xs
plot.respvspred Plot the relationship between Ys and Xs
plot.xpred maps the predictors contained in XXXpred
plot.correlation Calculate the correlations between Xs and eliminate Xs from candidate predictors if maximum correlation has been set with corlim
corlim Maximum correlation allowed between predictors, if higher correlation found predictors are withdrawn from Xs candidate list. If correlated, the last selected Xs is removed
use.correlation Select a model using only the uncorrelated predictors selected earlier
reset.cor resetting candidate predictors (Xs) to all selected Xs
full.models Select a model for each Ys forcing all candidate Xs in the models
stepwise.models Select a model for each Ys from candidate Xs using the different possible options below
test Choice of criterias (AIC, BIC) or tests (CHI and F (for quasi models))
plimit Probability limits for CHI and F tests
direction Set the direction of selection, default is both
df1 Optional smoother degree of freedom to be tested in variable selection
df2 First smoother degree of freedom to be tested in variable selection
startwith Vector of column numbers of predictors used to start stepwise procedure
resetstart Resetting starting predictors (Xs) to all selected Xs
contributions Calculate contributions of univariate models (alone), when dropping each predictor from the selected model (drop) and within the model (model)
plot.contributions Plot the calculated contributions as bars
plot.models Plot response curves of the selected models (using linear predictor scale before transformation into the response scale)
response.curves Plot combined response curves of the selected models (using response scale after transformation by the inverse link function)
model.anova Calculate an ANOVA table by removing each term from the full model
validate.models Validate the selected models for each Ys using both simple validation and cross-validation
predictions Calculate predictions for each Ys using selected model and new data (XXXpred) and save them in gr.predmat
export.predictions Export saved predictions from gr.predmat to a text file that can be directly imported in Arcview as an ASCII GRID
plot.predictions Plot saved predictions from gr.predmat
lookup.tables Calculate lookup tables describing partial responses curves at the linear predictor scale for each Ys using minimum and maximum values for each Xs contained in XXXpred. These values are derived from min and max values of the GRIDs that are used to build spatial predictions in Arcview.
plotpar number of rows and columns for multiple graphical outputs
prescol Color for presences in graph outputs
abscol Color for absences in graph outputs
nbar Number of bars in histograms
show.details Verbose output of stepwise selection
StdError Show pointwise error around partial plot
cvgroups Number of groups for cross validation
resolution Resolution for gids maps

Value

MODELCALLS Calls of selected models
VALIDATION Validation statistics for selected models
ALONE.CONTRIB Predictor alone contributions for selected models
DROP.CONTRIB Predictor drop contributions for selected models
MODEL.CONTRIB Predictor model contributions for selected models

Author(s)

Anthony.Lehmann@unige.ch

References

  • A. Lehmann, J.R. Leathwick & J.McC. Overton, 2002. GRASP. Ecological Modelling, 157: 189-207
  • HOMEPAGE: http://www.unige.ch/ia/climate/grasp
  • RELATED: http://www.spacedoor.net

    See Also

    grasp.in

    Examples

    
    data(YYY) # reads in YYY,XXX and XXXpred demo dataset
    data(XXX)
    data(XXXpred)
    grasp.in(YYY,XXX,XXXpred) # initialize a new grasp session
    
    grasp(2:3,c(4:6,8:9), title = "GRASP: ", path = "", gr.fam = "binomial", weights = TRUE, make.summary = TRUE, plot.maps = TRUE, plot.distry = TRUE, plot.histograms = TRUE, plot.respvspred = TRUE, plot.xpred = TRUE,plot.correlation = TRUE, stepwise.models = TRUE, test = "AIC", contributions = TRUE, plot.contributions = TRUE, plot.models = TRUE, model.anova = TRUE, validate.models = TRUE, predictions = TRUE, plot.predictions = TRUE)
    # run a full grasp analysis on 2 responses with 5 predictors
    
    

    [Package grasp version 2.5-7 Index]