Changes in 0.6-1
* Vignette and schoolarithm.Rd: fix typos
Changes in 0.6-0
* DESCRIPTION: suggests relations and Rgraphviz for plotting knowledge
structures and precedence relations, see examples in is.subset() and
in ita()
* New vignette: vignette("EstimationStepByStep", package = "pks"), parameter
estimation methods implemented in the package
* New function: slm(), fitting and testing simple learning models
* New function: getKFringe(), outer and inner fringes of a knowledge
structure
* New function: is.downgradable(), check for downgradabiliy
* conversion.R: as.pattern(..., useNames = TRUE) uses item names to
represent response patterns, as.letters argument is deprecated
* ita.R: ita() gains search argument for local or global threshold search;
returns precedence relation as a logical incidence matrix
* New function: is.subset(), subset relation incidence matrix
* New function: blimit(), BLIM identification analysis (adapted from
Stefanutti et al., 2012)
* New data sets: fraction17 (Stefanutti and de Chiusole, 2017),
subtraction13 (de Chiusole and Stefanutti, 2013)
Changes in 0.5-0
* blim(): use (if possible) matrix multiplication to compute the conditional
distribution of response patterns given knowledge state; remove deprecated
errequal and errtype arguments; increase default zeropad to 16
* New function: ita(), item tree analysis
* conversion.R: as.binmat() gains as.logical argument
* chess data: correct mistakes in dst1 and dst4; add raw data and
replication code
* *.Rd: update references
Changes in 0.4-1
* BUG FIX: blim(..., method = "MDML") could fail when (b)eta estimates were
too close to zero (reported by Martin Losert)
* *.Rd: add DOIs to all references
* DESCRIPTION: new URL
Changes in 0.4-0
* blim(): more flexible infrastructure for setting parameter constraints;
errequal and errtype arguments are now deprecated in favor of
betaequal/etaequal and betafix/etafix
* New functions: anova.blim(), coef.blim(), deviance.blim()
* as.pattern(..., as.set = TRUE), as.binmat(): work in more cases including
when R or N.R do not contain the empty set
* chess and probability data sets: extended examples
* NAMESPACE: import graphics
Changes in 0.3-1
* BUG FIX: delineate() could return wrong knowledge structure when skillfun
had numeric itemIDs (or factor itemIDs in non-alphabetic order)
* New function: jacobian(), Jacobian matrix of a BLIM (draft version)
* delineate(): each element of classes is now a matrix; itemID is now the
index of the column that holds the item indicator
* blim(): new zeropad argument; per default, an incomplete response vector
is completed and padded with zeros
Changes in 0.3-0
* New function: delineate(), knowledge structure delineated by skill
function
* New functions: is.forward.graded(), is.backward.graded()
* blim(): add random initial values for randinit = TRUE
* tests/sampling.R: add tests for uniform sampling
* blim(): recalculate degrees of freedom in goodness-of-fit test
* as.pattern(): add as.set argument; if TRUE return value is of class "set"
* as.binmat(): conversion from class set to matrix
* DESCRIPTION: depend on package sets
Changes in 0.2-1
* BUG FIX: simulate.blim() made incorrect predictions
* New data set: endm
* CITATION: add reference to ENDM paper
Changes in 0.2-0
* BUG FIX: likelihood and fitted values reported by blim() were incorrect
* blim(): add estimators for errequal = TRUE, speed up some computations
* New data sets: density97, matter97
* as.pattern(): add as.letters argument for a letter representation of
response patterns
* CITATION: add reference to Psychoco 2012 talk
Changes in 0.1-0
* initial release on CRAN
* Function blim() for fitting and testing basic local independence models
* Extractor functions for blim objects (some work automatically by
inheritance): simulate(), print(), logLik(), nobs(), resid(), fitted(),
plot(), AIC(), BIC()
* Conversion functions: as.pattern() and as.binmat()
* Data sets: DoignonFalmagne7, chess, probability