#48 Correct interpretation of log-normal distribution

[Anonymous]

The most recent JAGS user manual is misleading as to how the lognormal distribution is defined. On page 46 of the manual it states

"As the name suggests, the log-normal distribution is the log transformation of a normal
distribution with mean µ and precision τ . It is equivalent to
y <- log(x)
x ~ dnorm(mu, tau)"

However, this is backwards - if X ~ LOGNOR(mu,sigma) and Y = log(X) then Y ~ NORM(mu,sigma). Taking the log of normally distributed data does not result in any known distribution that I'm aware of. You code should read

y <- exp(x)
x ~ dnorm(mu, tau)

Further complicating this issue is that the density function shown on page 46 for the lognormal distribution is correct. I wanted to confirm that the lognormal distribution is being implemented correctly within JAGS