fitslikDT: Maximum likelihood fits for psychometric functions
In VinLaflamme/nlregVL: Non-linear fits for psychometric functions

Maximum likelihood fits for psychometric functions

fitslikDT(dataF, cD, cNbS, cNbL, cBy, type = "CG", getPred, loglikfn,
  loglikgr, loglikgr2, fnStart, method = "BFGS", maxN = 100,
  parstart.allow = F, allow.print = F, ctrlL = list(fnscale = -1, reltol =
  .Machine$double.eps))

`dataF`	The data table containing the data to be fitted
`cD`	name (or position) of the column containing the x-axis data. In the case of a temporal bisection task, this corresponds to the name (or position) of the column containing the duration.
`cNbS`	name (or position) of the column containing the number of "short" responses.
`cBy`	name (or position) of the column(s) to group by. If left empty, it will take all other columns.
`type`	string expliciting which of three available fitting functions will be used. "CG" for cumulative Gaussian, "LN" for lognormal and "PL" for pseudologistic.
`getPred`	Function that returns the predicted value.
`loglikfn`	Only used if type is not specified. A function that computes the loglikelihood. It must take the following arguments: par (a vector of numerical values), Dur (a vector of durations), NbS (vector of the number of short respnses), NbL (vector of the number of short respnses). It must return a numerical vector of size 1.
`loglikgr`	Only used if type is not specified. A function that computes the gradient of the loglikelihood. It must take the following arguments: par (a vector of numerical values), Dur (a vector of durations), NbS (vector of the number of short respnses), NbL (vector of the number of short respnses). It must return a numerical vector of the same size as par.
`loglikgr2`	Only used if type is not specified. A function that computes the second order gradient of the loglikelihood. It must take the following arguments: par (a vector of numerical values), Dur (a vector of durations), NbS (vector of the number of short respnses), NbL (vector of the number of short respnses). It must return a numerical vector of the same size as par.
`fnStart`	A function that calculates the starting values for the fitting process. It must take a single argument (Dur).
`method`	See `optim`
`maxN`	Maximum number of iterations.
`parstart.allow`	If the starting parameters are bad, allows the algorithm to find new ones.
`allow.print`	Allows the printing of the number of iterations of the fits.
`ctrlL`	control list for optim. See `optim`

A data.table containing the parameter estimates, the standard error of those estimates, the log likelihood as well as an information code (see optim)

VinLaflamme/nlregVL documentation built on May 30, 2019, 2:04 a.m.