What is the interpretation of the lam parameter in StudentT distribution? I am using a bayesian linear regression with StudentT as the prior. In the result, when I get the mean coefficients for all the features, there is one value for lam also. How to interpret this?
It’s the degrees of freedom of the t distribution. When the degrees of freedom are very large, then the distribution is equal to a gaussian. When lam
is very small, then the tails carry much more weight
1 Like
According to the documentation, nu is the degrees of freedom. Lam is the scale parameter. What is the interpretation for scale parameter?
1 Like
Oops. Sorry for the mix-up. nu
are the degrees of freedom and lam
is the precision of the gaussian to which the t distribution converges when nu
gets really big. The precision is the inverse of the variance of the gaussian. So the bigger the precision, the narrower the distribution is
1 Like