Given a dark object, obj, this function repeatedly optimises the parameters in the vicinity of the seed array. The width of the search is dependent upon the value of spread.
Usage
MultiStart(obj, repeats, draw, spread, debug)
Arguments
obj
A dark object containing at least;ll{
obj$time time
obj$thrs thresholds
obj$init an initial estimate of the parameters of dark adaptation.
}
repeats
The number of times the algorithm is repeated
draw
A flag indicating whether a figure should be drawn.
spread
The amount by which the seed array should be varied. A larger value gives a greater range of possible starting points.
debug
A flag used in debugging the software.
Value
Returns a list;
timetimes of threshold setting
out$thrsobserved thresholds
out$residresiduals
out$fitoptimal fitted values
out$thetseed parameters if test data
out$ssesum of squared residuals if test data
out$datasource of the data
out$optoptimal parameter estimates of the chosen model
out$Modname of the optimal model
out$Pnthe number of parameters needed to describe the data
out$AICarray of AICc scores
out$valcalculated sum of squared residuals
out$R2the coefficient of determination
out$warningif none of the nearby values converge
out$callupdates the function call label
Details
To reduce the possibility of selecting non-optimal parameter estimates, the optimisation is repeated in the region of initial estimates. The
References
Nelder, J.A.; Mead, R. 1965: A simplex for function minimization. Comput. J. 7, 308-313