Statscraft 0.15.0, release v0.15.0
BAYES CORRELATION
Bayesian inference on the Pearson correlation of every pair of variables.
Syntax
BAYES CORRELATION /INFERENCE VARIABLES=varlist [ANALYSIS=POSTERIOR | BAYESFACTOR | BOTH]
[CVALUE=0] [MAXPLOTS=10]
[/ESTBF [TYPE=JZS | FRACTION[(0.5)]] [NULLVALUE=0]]
[/CRITERIA [CILEVEL=95] [SEED=RANDOM | n] [TOL=0.0001] [MAXITER=2000] [MCSAMPLES=10000]
[POSTSAMPLES=10000]] [/MISSING SCOPE=PAIRWISE | LISTWISE].
Every variable is paired with every later one. JZS gives BF01 of a zero correlation
against the default prior of the linear model; FRACTION(b) the fractional Bayes factor
BF01 of NULLVALUE, with the fraction b of the data training the prior (b times N must
exceed 1). A correlation of 1 or -1 gets no Bayes factor. The posterior of rho is
proportional to (1 - rho^2)^CVALUE times the approximate likelihood; its mode, mean,
variance and highest-density interval are computed by quadrature, so SEED and
POSTSAMPLES do not change them. Unless ANALYSIS=BAYESFACTOR, the first MAXPLOTS pairs get
a chart of the prior, likelihood and posterior densities of rho as chart data. MCSAMPLES is
accepted and not used. Weights as on BAYES ONESAMPLE.
Example
BAYES CORRELATION /INFERENCE VARIABLES=sleep_hours focus_score mood ANALYSIS=BOTH
/ESTBF TYPE=FRACTION(0.4).
statscraft help BAYES CORRELATION prints the same text in the terminal. All commands