Bayesian Correlation using PyMC -

i'm pretty new pymc, , i'm trying implement simple bayesian correlation model, defined in chapter 5 of "bayesian cognitive modeling: practical course", defined below:

i've put code in ipython notebook here, code snippet follows:

mu1 = normal('mu1', 0, 0.001) mu2 = normal('mu2', 0, 0.001) lambda1 = gamma('lambda1', 0.001, 0.001) lambda2 = gamma('lambda2', 0.001, 0.001) rho = uniform('r', -1, 1)  @pymc.deterministic def mean(mu1=mu1, mu2=mu2):     return np.array([mu1, mu2])  @pymc.deterministic def precision(lambda1=lambda1, lambda2=lambda2, rho=rho):     sigma1 = 1 / sqrt(lambda1)     sigma2 = 1 / sqrt(lambda2)     ss1 = sigma1 * sigma2     ss2 = sigma2 * sigma2     rss = rho * sigma1 * sigma2     return np.power(np.mat([[ss1, rss], [rss, ss2]]), -1)  xy = mvnormal('xy', mu=mean, tau=precision, value=data, observed=true)  m = pymc.mcmc(locals()) m.sample(10000, 5000) 

the error "error: failed in converting 3rd argument `tau' of flib.prec_mvnorm c/fortran array"

i found 1 other reference error (in this question) couldn't see how apply answer there code.

this uninformative error due way have organized data vector. 2 rows n columns, , pymc expects n rows 2 columns. following modification makes code (almost) work me:

xy = mvnormal('xy', mu=mean, tau=precision, value=data.t, observed=true) 

i because changed precision matrix not have matrix power part. think mvnormal in figure has variance-covariance matrix second parameter, while mvnormal in pymc expects precision matrix (equal inverse of c).

here notebook has no more errors, has warning requires additional investigation.