Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Statscraft 0.15.0, release v0.15.0

BAYES LOGLINEAR

Bayesian test of independence in a two-way table, with simulated interaction terms.

Syntax

BAYES LOGLINEAR /TABLE rowvar BY colvar
   [/INFERENCE ANALYSIS=POSTERIOR | BAYESFACTOR | BOTH]
   [/ESTBF [MODEL=MULTINOMIAL | POISSON | NONPARAMETRIC] [FIX_MARGIN=TOTAL | ROW | COLUMN]
           [PRIOR_TYPE=CONJUGATE | MIXTURE_DIRICHLET | INTRINSIC] [SHAPEPARAMS=values]
           [SCALEPARAM=DATA | b] [LAMBDAS=values]]
   [/PRINT [SUPPRESS_TABLE=FALSE | TRUE] [CELLS=OBSERVED ROW COLUMN EXPECTED TOTAL ALL]
           [STATISTICS=CHISQ LRT]]
   [/CRITERIA [CILEVEL=95] [SEED=RANDOM | n] [TOL=0.000001] [MAXITER=2000]
              [POSTSAMPLES=10000] [ORDER=ASCENDING | DESCENDING]].
The table counts cases valid on both variables with rounded case weights; only categories
that occur count, and ORDER sorts both of them. Every Bayes factor is BF01, for
independence. CONJUGATE puts gamma priors with the shapes of SHAPEPARAMS (1 for every cell;
one value, or one row of values per table row with ';' between rows) on the cells;
FIX_MARGIN says whether the grand total, the row totals or the column totals are fixed by
the design. POISSON fixes nothing and uses the gamma scale SCALEPARAM (DATA: rows times
columns times the smallest shape over N). MIXTURE_DIRICHLET mixes symmetric Dirichlet priors
over a log-Cauchy weight, by quadrature with TOL and MAXITER. INTRINSIC needs a 2 x 2 table
(others fall back to the mixture with a warning). NONPARAMETRIC needs FIX_MARGIN=ROW with
two rows or COLUMN with two columns; LAMBDAS holds one Dirichlet parameter, or one per free
category.
The posterior part draws POSTSAMPLES cell probability sets from SEED and reports each log
odds ratio against the last row and column, with a simultaneous credible region for all of
them. PRINT shows the crosstab (observed counts by default) and, for CHISQ, Pearson's
chi-square, the continuity correction of a 2 x 2 table and Fisher's test when an expected count
is below 5; LRT adds the likelihood ratio. Keywords with an underscore may be written
without it or as two words.

Example

BAYES LOGLINEAR /TABLE ferry_route BY ticket_class
/INFERENCE ANALYSIS=BOTH /ESTBF MODEL=MULTINOMIAL FIX_MARGIN=ROW PRIOR_TYPE=CONJUGATE
/PRINT CELLS=OBSERVED ROW STATISTICS=CHISQ LRT /CRITERIA SEED=4417.

statscraft help BAYES LOGLINEAR prints the same text in the terminal. All commands