Given response data that form a connected design, compute the probability to pass on the reference set conditional on each score on one or more target tests.
probability_to_pass(dataSrc, ref_items, pass_fail, design = NULL,
predicate = NULL, similar_groups = TRUE)Data source: a dexter project db handle or a data.frame with columns: person_id, item_id, item_score and booklet_id
vector with id's of items in the reference set, they must all occur in dataSrc
pass-fail score on the reference set
A data.frame with columns booklet_id (if multiple booklets) and item_id defining the target test booklet(s), if NULL (default) this will be derived from the dataSrc and the probability to pass will be computed for each test score for each booklet in your data.
An optional expression to subset data in dataSrc, if NULL all data is used
When TRUE it is assumed that candidates taking the reference test and the target tests are similar in ability. If they are similar, the score distribution on the target test is estimated more precisely. Works only when design contains booklet_id's.
An object of type p2pass. Use coef() to extract the
probablity to pass for each booklet and score. Use plot() to plot
the probabilities, sensitivity and specificity.
Note that this function is computationally intensive and can take a long time to run, especially when computing the probability to pass for multiple target booklets.
For any possible score k, we use a Gibbs sampler to calculate
where:
is student ability
are the item parameters
is the score on the reference test
is an established pass_fail score on the reference test
is the score on the booklet
are the observed data
This probability can be used to establish a pass-fail score for the new booklet.