gLVnonlinearRegression: Parameter estimation through gradient search of continuous...

Usage Arguments Value

Usage

1
gLVnonlinearRegression(data, parms0 = NULL, ftol = 1e-8 , ptol = 1e-8, maxiter = 100, lowerbound = rep(-20,ncol(data$obs)*(ncol(data$obs)-1)), upperbound = rep(20,ncol(data$obs)*(ncol(data$obs)-1)), method = "Marq")

Arguments

data

Data input containing a time series of observations in longitudinal matrix form

parms0

Optional. Starting parameter vector. Default = Zero vector

ftol

Minimum change in the value of the objective function (sum of squared residuals) between two consecutive steps before stopping iterative optimization

ptol

Minimum change in the value of parameters being estimated between two consecutive steps before stopping iterative optimization

maxiter

Maximal number of iterations allowed before breaking the optimization algorithm

lowerbound

Numerical vector of equal length as parameter vector describing lower bound for constrained parameter search

upperbound

Numerical vector of equal length as parameter vector describing upper bound for constrained parameter search

method

Method used for optimization of the objective function. Default is "Marq" for Leverberg-Marquandt

Value

Parms

Estimated parameter matrix

SSR

The value of the sum of squared residuals of the found solution

residual_SD

The estimated standard deviation of the data stochasticity in respect to the solution found

SE

A matrix contatining the standard errors of each estimated parameter

residuals_t.test

The result of a t test with the null hypothesis of the residuals from the model solution being normally distributed with standard error 'residual_SD'

message

A text message related to the reason why the optimization algorithm stopped

obs

A matrix with the calculated abundances on each timestep according to the estimated parameters

Fit

The object returned by the optimization algorithm. It is a list containing further details of the optimization algorithm run

df

The degrees of freedom

quantitative

Only when validating against in Silico data: The ratio of estimated parameters that contain the original value within the 95% confidence interval given by their standard error (displayed in 'SE')

qualitative1

Only when validating against in Silico data: The ratio of correctly retrieved edges from the original interaction network

qualititative2

Only when validating against in Silico data: The ratio of correctly retrieved edges from the original interaction network after setting estimated parameters that are not significantly different from zero to zero (i.e. parameters that contain the zero in their 95% confidence interval as given by their standard error)


lkshrsch/gLVInterNetworks documentation built on May 21, 2019, 7:33 a.m.