diff --git a/scvi/module/_amortizedlda.py b/scvi/module/_amortizedlda.py
index d1389f564d..5f3f63b513 100644
--- a/scvi/module/_amortizedlda.py
+++ b/scvi/module/_amortizedlda.py
@@ -30,9 +30,15 @@ def log_prob(self, value):
 def logistic_normal_approximation(
     alpha: torch.Tensor,
 ) -> Tuple[torch.Tensor, torch.Tensor]:
+    """
+    Returns the mean and standard deviation of the Logistic Normal approximation to the Dirichlet.
+
+    Uses the Laplace approximation of the Logistic Normal distribution to the Dirichlet distribution
+    as described in Srivastava et al. https://arxiv.org/pdf/1703.01488.pdf.
+    """
     K = alpha.shape[0]
     mu = torch.log(alpha) - torch.log(alpha).sum() / K
-    sigma = (1 - 2 / K) / alpha + torch.sum(1 / alpha) / K ** 2
+    sigma = torch.sqrt((1 - 2 / K) / alpha + torch.sum(1 / alpha) / K ** 2)
     return mu, sigma