xxirt.Rd
Estimates a user defined item response model. Both, item response functions and latent trait distributions can be specified by the user (see Details).
xxirt(dat, Theta=NULL, itemtype=NULL, customItems=NULL, partable=NULL, customTheta=NULL, group=NULL, weights=NULL, globconv=1e-06, conv=1e-04, maxit=1000, mstep_iter=4, mstep_reltol=1e-06, h=1E-4, use_grad=TRUE, verbose=TRUE, penalty_fun_item=NULL, np_fun_item=NULL, verbose_index=NULL, cv_kfold=0, cv_maxit=10) # S3 method for xxirt summary(object, digits=3, file=NULL, ...) # S3 method for xxirt print(x, ...) # S3 method for xxirt anova(object,...) # S3 method for xxirt coef(object,...) # S3 method for xxirt logLik(object,...) # S3 method for xxirt vcov(object,...) # S3 method for xxirt confint(object, parm, level=.95, ... ) # S3 method for xxirt IRT.expectedCounts(object,...) # S3 method for xxirt IRT.factor.scores(object, type="EAP", ...) # S3 method for xxirt IRT.irfprob(object,...) # S3 method for xxirt IRT.likelihood(object,...) # S3 method for xxirt IRT.posterior(object,...) # S3 method for xxirt IRT.modelfit(object,...) # S3 method for IRT.modelfit.xxirt summary(object,...) # S3 method for xxirt IRT.se(object,...) # computes Hessian matrix xxirt_hessian(object)
dat | Data frame with item responses |
---|---|
Theta | Matrix with \(\bold{\theta}\) grid vector of latent trait |
itemtype | Vector of item types |
customItems | List containing types of item response functions created by
|
partable | Item parameter table which is initially created by
|
customTheta | User defined \(\bold{\theta}\) distribution created by
|
group | Optional vector of group indicators |
weights | Optional vector of person weights |
globconv | Convergence criterion for relative change in deviance |
conv | Convergence criterion for absolute change in parameters |
maxit | Maximum number of iterations |
mstep_iter | Maximum number of iterations in M-step |
mstep_reltol | Convergence criterion in M-step |
h | Numerical differentiation parameter |
use_grad | Logical indicating whether the gradient should be supplied
to |
verbose | Logical indicating whether iteration progress should be displayed |
penalty_fun_item | Optional penalty function used in regularized
estimation. Used as a function of |
np_fun_item | Function that counts the number of item parameters in regularized
estimation. Used as a function of |
object | Object of class |
digits | Number of digits to be rounded |
file | Optional file name to which |
parm | Optional vector of parameters |
level | Confidence level |
verbose_index | Logical indicating whether item index should be printed in estimation output |
cv_kfold | Number of k folds in cross validation. The default is 0 (no cross-validation) |
cv_maxit | Maximum number of iterations for each cross-validation sample |
x | Object of class |
type | Type of person parameter estimate. Currently, only
|
... | Further arguments to be passed |
Item response functions can be specified as functions of unknown parameters \(\bold{\delta}_i\) such that \(P(X_{i}=x | \bold{\theta})=f_i( x | \bold{\theta} ; \bold{\delta}_i )\) The item response model is estimated under the assumption of local stochastic independence of items. Equality constraints of item parameters \(\bold{\delta}_i\) among items are allowed.
The probability distribution \(P(\bold{\theta})\) are specified as functions of an unknown parameter vector \(\bold{\gamma}\).
A penalty function for item parameters can be specified in
penalty_fun_item
. The penalty function should be differentiable and
a non-differentiable function (e.g., the absolute value function) should
be approximated by a differentiable function.
List with following entries
Item parameter table
Vector with estimated item parameters
Data frame with item parameters
Data frame with summary on bounds of estimated item parameters
Vector with estimated parameters of theta distribution
Matrix with \(\bold{\theta}\) grid
Item response functions
Theta distribution
Deviance
Log likelihood value
Information criteria
List with item functions
Used customized item response functions
Used customized theta distribution
Cross-validated log-likelihood value (if cv_kfold>0
)
Individual likelihood
Individual posterior
Case-wise log-likelihood values
Array of expected counts
EAP person parameter estimates
Used dataset with item responses
Dataset with response indicators
Vector of person weights
Number of groups
Integer vector of group indicators
Vector of original group_identifiers
Number of categories per item
Logical whether model has converged
Number of iterations needed
See the mirt::createItem
and
mirt::mirt
functions in the mirt
package for similar functionality.