--- a/src/modules/glm/samplers/ConjugateFMethod.h
+++ b/src/modules/glm/samplers/ConjugateFMethod.h
@@ -4,21 +4,35 @@
 #include <sampler/SampleMethod.h>
 #include <sampler/GraphView.h>
 
-/**
- * When an F distribution is used as a prior for the precision of
- * normally distributed random effects, it has a conditionally
- * conjugate distribution using a redundant parametrization.
- */
-class ConjugateFMethod {
-    GraphView *_gv1, *_gv2;
-    unsigned int _chain;
-    double _scale, _tau;
-    double *_coef;
-public:
-    ConjugateFMethod(GraphView *gv1, GraphView *gv2, unsigned int chain);
-    ~ConjugateFMethod();
-    void update(RNG *rng);
-    static bool canSample(StochasticNode *snode, Graph const &graph);
-};
+namespace glm {
+
+    /**
+     * @short Sampling method for precision parameters of random effects.
+     *
+     * When an F distribution is used as a prior for the precision of
+     * normally distributed random effects in a linear model, it has a
+     * conditionally conjugate distribution using a redundant
+     * parametrization.
+     *
+     * This is described by Gelman A (2006) Prior distributions for
+     * variance parameters in hierarchical models. Bayesian Analysis
+     * 1:515���533. Gelman describes the half-t distribution of the standard
+     * deviation. This corresponds to an F(m,1) prior on the precision
+     * parameter.
+     *
+     * This is currently an experimental sampling method.
+     */
+    class ConjugateFMethod {
+	GraphView *_gv1, *_gv2;
+	unsigned int _chain;
+	double _scale, _tau;
+	double *_coef;
+    public:
+	ConjugateFMethod(GraphView *gv1, GraphView *gv2, unsigned int chain);
+	~ConjugateFMethod();
+	void update(RNG *rng);
+    };
+
+}
 
 #endif /* CONJUGATE_F_METHOD_H_ */