Extending conditional likelihood for clustered data