A random‐effects model for clustered circular data