DPSD: Wrapper for fitting the dual-process model to recognition and...

Description Usage Arguments Details Value Author(s) References

View source: R/DPSD.R

Description

This function fits the univariate DPSD model (Yonelinas, 1999) to recognition or source memory item-level rating data.

Usage

1
2
3
DPSD(responseScale, confidenceRatings, TargetLure, targetLabel, lureLabel,
  iterations = 200, eqVar = TRUE, eqRecollection = FALSE,
  fitSource = TRUE, fitROC = FALSE)

Arguments

responseScale

An vector containing possible levels of confidence rating responses ordered from highest to lowest (e.g. 6:1).

confidenceRatings

An vector containing participant confidence rating responses to individual items.

TargetLure

An vector coding containing information whether each individual item is a Target or Lure item (e.g. Old, New, Old, Old)

targetLabel

A string/integer designating the label for the target items in TargetLure

lureLabel

A string/integer designating the label for the lure items in TargetLure

iterations

A numeric value specifying the number of iterations. Default is set to 200.

eqVar

A boolean value specifying if the standard deviation of the target distribution is equal to that of the lure distribution (i.e. = 1) (TRUE) or estimated separately (FALSE). Default is set to TRUE.

eqRecollection

A boolean value specifying if recollection is set equal for the target and lure source (TRUE) or is estimates separately for both sources (FALSE). Default is set to FALSE.

fitSource

A boolean value specifying if the model is being fitted to source memory data (TRUE) or recognition memory data (FALSE). Default is set to TRUE

fitROC

A boolean specifying if the model is fitted by minimizing SSE to ROCs or by MLE to rating frequencies

Details

The optimization can be attempted by by minimizing the total squared difference (SSE) between observed and predicted cumulative hit and false alarm rates by setting "fitROC = TRUE" (default is fitROC = FALSE) or by minimzing the negative log-likelihood for rating frequencies by setting "fitROC = FALSE". In both cases, fitting is done using the Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm in optim.

When setting fitROC = TRUE, the data is transformed into cumulative hit/fa rates (via cumRates.R) prior to fitting. When setting fitROC = FALSE, the data is transformed into frequency data (via ratingFreq.R) prior to fitting.

Arguments in the function allow the user to specify if they are fitting recognition or source memory data, an equal- or unequal-variance model and when fitting source memory data, if a single recollection parameter equal for both sources or two separate recollection parameters are to be estimated.

Recollection is bounded to be between 0 and 1, Familiarity and the standard deviation of the target distribution to be positive. Criteria are unbounded, for MLE, criteria are ordered.

Each model is fit to each data set for a set number of times (default is iterations = 200), and the iteration with the minimum minimized value is output.

Value

The function returns a dataframe with components:

(parameters)

The estimated parameters (recollection_target, recollection_lure, familiarity, sd_target, criteria) for the iteration with the lowest SSE or likelihood

SSE if fitROC = TRUE

Minimum sum square error

negLL if fitROC = FALSE

negative Log Likelihood

Author(s)

Nicholas Lange, lange.nk@gmail.com

References

Yonelinas, A. P. (1999). The Contribution of Recollection and Familiarity to Recognition and Source-Memory Judgments: A Formal Dual-Process Model and an Analysis of Receiver Operating Characteristics. Journal of Experimental Psychology: Learning, Memory, and Cognition, 25(6), 1415 - 1434. http://doi.org/10.1037//0278-7393.25.6.1415


nklange/fullDPROC documentation built on May 26, 2019, 2:34 a.m.