fix per comments

arviz/plots/khatplot.py

@@ -180,7 +180,13 @@ def plot_khat(
             good_k = khats.good_k
             khats = khats.pareto_k
         if not isinstance(khats, DataArray):
-            raise ValueError("Incorrect khat data input. Check the documentation")
+            good_k = None
+            warnings.warn(
+                "support for DataArrays will be deprecated, please use ELPDData."
+                "The reason for this, is that we need to know the numbers of draws"
+                "sampled from the posterior",
+                FutureWarning,
+            )
 
         khats = get_coords(khats, coords)
         dims = khats.dims

arviz/stats/stats.py

@@ -790,7 +790,7 @@ def loo(data, pointwise=None, var_name=None, reff=None, scale=None):
 
     if np.any(pareto_shape > good_k):
         warnings.warn(
-            f"Estimated shape parameter of Pareto distribution is greater than {good_k:.1f} "
+            f"Estimated shape parameter of Pareto distribution is greater than {good_k:.2f} "
             "for one or more samples. You should consider using a more robust model, this is "
             "because importance sampling is less likely to work well if the marginal posterior "
             "and LOO posterior are very different. This is more likely to happen with a "

arviz/stats/stats_utils.py

@@ -454,8 +454,8 @@ def get_log_likelihood(idata, var_name=None, single_var=True):
 
 Pareto k diagnostic values:
                          {{0:>{0}}} {{1:>6}}
-(-Inf, {{8:.1f}}]   (good)     {{2:{0}d}} {{5:6.1f}}%
-   ({{8:.1f}}, 1]   (bad)      {{3:{0}d}} {{6:6.1f}}%
+(-Inf, {{8:.2f}}]   (good)     {{2:{0}d}} {{5:6.1f}}%
+   ({{8:.2f}}, 1]   (bad)      {{3:{0}d}} {{6:6.1f}}%
    (1, Inf)   (very bad) {{4:{0}d}} {{7:6.1f}}%
 """
 SCALE_DICT = {"deviance": "deviance", "log": "elpd", "negative_log": "-elpd"}