View source: R/count_binomial.R
count_binom | R Documentation |
Draws prior/posterior samples for product-binomial data and counts how many samples are
inside the convex polytope defined by
(1) the inequalities A*x <= b
or
(2) the convex hull over the vertices V
.
count_binom( k, n, A, b, V, map, prior = c(1, 1), M = 10000, steps, start, cmin = 0, maxiter = 500, burnin = 5, progress = TRUE, cpu = 1 )
k |
vector of observed response frequencies. |
n |
the number of choices per item type.
If |
A |
a matrix with one row for each linear inequality constraint and one
column for each of the free parameters. The parameter space is defined
as all probabilities |
b |
a vector of the same length as the number of rows of |
V |
a matrix of vertices (one per row) that define the polytope of
admissible parameters as the convex hull over these points
(if provided, |
map |
optional: numeric vector of the same length as |
prior |
a vector with two positive numbers defining the shape parameters of the beta prior distributions for each binomial rate parameter. |
M |
number of posterior samples drawn from the encompassing model |
steps |
an integer vector that indicates the row numbers at which the matrix |
start |
only relevant if |
cmin |
if |
maxiter |
if |
burnin |
number of burnin samples per step that are discarded. Since the
maximum-likelihood estimate is used as a start value (which is updated for each step in
the stepwise procedure in |
progress |
whether a progress bar should be shown (if |
cpu |
either the number of CPUs used for parallel sampling, or a parallel
cluster (e.g., |
Counts the number of random samples drawn from beta distributions that satisfy
the convex, linear-inequalitiy constraints. The function is useful to compute
the encompassing Bayes factor for testing inequality-constrained models
(see bf_binom
; Hojtink, 2011).
The stepwise computation of the Bayes factor proceeds as follows:
If the steps are defined as steps=c(5,10)
, the BF is computed in three steps by comparing:
(1) Unconstrained model vs. inequalities in A[1:5,]
;
(2) use posterior based on inequalities in A[1:5,]
and check inequalities A[6:10,]
;
(3) sample from A[1:10,] and check inequalities in A[11:nrow(A),]
(i.e., all inequalities).
a matrix with the columns
count
: number of samples in polytope / that satisfy order constraints
M
: the total number of samples in each step
steps
: the "steps"
used to sample from the polytope
(i.e., the row numbers of A
that were considered stepwise)
with the attributes:
proportion
: estimated probability that samples are in polytope
se
: standard error of probability estimate
const_map
: logarithm of the binomial constants that
have to be considered due to equality constraints
Hoijtink, H. (2011). Informative Hypotheses: Theory and Practice for Behavioral and Social Scientists. Boca Raton, FL: Chapman & Hall/CRC.
Fukuda, K. (2004). Is there an efficient way of determining whether a given point q is in the convex hull of a given finite set S of points in Rd? Retrieved from https://www.cs.mcgill.ca/~fukuda/soft/polyfaq/node22.html
bf_binom
, count_multinom
### a set of linear order constraints: ### x1 < x2 < .... < x6 < .50 A <- matrix( c( 1, -1, 0, 0, 0, 0, 0, 1, -1, 0, 0, 0, 0, 0, 1, -1, 0, 0, 0, 0, 0, 1, -1, 0, 0, 0, 0, 0, 1, -1, 0, 0, 0, 0, 0, 1 ), ncol = 6, byrow = TRUE ) b <- c(0, 0, 0, 0, 0, .50) ### check whether a specific vector is inside the polytope: A %*% c(.05, .1, .12, .16, .19, .23) <= b ### observed frequencies and number of observations: k <- c(0, 3, 2, 5, 3, 7) n <- rep(10, 6) ### count prior samples and compare to analytical result prior <- count_binom(0, 0, A, b, M = 5000, steps = 1:4) prior # to get the proportion: attr(prior, "proportion") (.50)^6 / factorial(6) ### count posterior samples + get Bayes factor posterior <- count_binom(k, n, A, b, M = 2000, steps = 1:4) count_to_bf(posterior, prior) ### automatic stepwise algorithm prior <- count_binom(0, 0, A, b, M = 500, cmin = 200) posterior <- count_binom(k, n, A, b, M = 500, cmin = 200) count_to_bf(posterior, prior)
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.