morris {sensitivity}R Documentation

The Morris OAT Screening Method

Description

morris is the implementation of the Morris OAT Screening method. This function generates the Morris design of experiments and computes the measures of sensitivity mu* and sigma.

Usage

morris(model = NULL, factors, levels, R, jump = NULL,
       min = 0, max = 1, scale = TRUE, optim = NULL, ...)

Arguments

model the model
factors the number of factors, or their names
levels the number of levels of the design grid
R the number of repetitions of the design, i.e. the number of elementary effect computed per factor
jump the ‘grid jump’ coefficient
min the minimum values for the factors
max the maximum values for the factors
scale logical. If TRUE, the input and output data are scaled
optim optimization of the design for better coverage of the space (cf Campolongo 2005), not documented yet (for informations feel free to ask the maintainer)
... any other arguments for model which are passed unchanged each time it is called

Details

model is a function or a predictor (a class with a predict method) computing the response y based on the sample given by x. If no model is specified, the indices will be computed when one gives the response.

factors could either be a single number or a vector of character strings.

The number of levels is not necessary the same for each space coordinate. It is the case when levels is a single integer.

min and max are boundaries of the region of experimentation. They can be single values (the same for each factor) or vectors.

jump is such that:

Delta[i] = jump[i] * ( max[i] - min[i] ) / ( levels[i] - 1 )

If jump is given as NULL and the number of levels is even (for each component), then jump has the value recommended by Morris: jump = levels / 2. If jump is a single value, then it is taken the same for each coordinate.

Value

morris returns an object of class "morris". An object of class "morris" is a list containing the following components:

x the design of experiments (input sample)
y the response
ee the matrix of the elementary effects
mu the estimations of the mu* index
sigma the estimations of the sigma index

Computational cost

The number of model evaluations is (p + 1) * R where p is the number of factors.

References

Saltelli, A., Chan, K. and Scott, E. M., 2000, Sensitivity analysis. Wiley.

Morris, M. D., 1991, Factorial sampling plans for preliminary computational experiments. Technometrics, 33, 161–174.

Examples

# Test case : the non-monotonic function of Morris

sa <- morris(model = morris.fun, factors = 20, levels = 4, R = 4)
print(sa)
plot(sa)

[Package sensitivity version 1.3-0 Index]