fixed a bug in trends that occurred randomly when using an incomplete time...

fixed a bug in trends that occurred randomly when using an incomplete time series, updated toarstats to 0.6.9

CHANGELOG.md

 # Changelog
 All notable changes to this project will be documented in this file.
 
+## v0.6.9 - 2025-04-04 - fixed bug in trends/moving_block_bootstrap
+
+### technical:
+* when calculating a trend on a timeseries which is not complete, the method
+`moving_block_bootstrap` may throw a `ValueError` based on randomly chosen blocks
+
 ## v0.6.8 - 2025-03-07 - bugfix in time harmonizing
 
 ## v0.6.7 - 2025-02-21 - fixed output format

setup.cfg

 [metadata]
 name = toarstats
-version = 0.6.8
+version = 0.6.9
 author = Niklas Selke, Martin Schultz, Max Lensing
 author_email = n.selke@fz-juelich.de, m.schultz@fz-juelich.de, m.lensing@fz-juelich.de
 description = Collection of statistics for the TOAR community

toarstats/trends/utils.py

@@ -63,9 +63,12 @@ def moving_block_bootstrap(method, data, quantile=None, num_samples=1000):
     for _ in range(num_samples):
         bn = rng.choice(len(blocks), nblocks)
         samp_data = data.iloc[blocks[bn].flatten()]
+        try:
             if method == "quant":
                 mod = smf.quantreg("value~datetime", samp_data).fit(q=quantile)
             else:
                 mod = smf.ols("value~datetime", samp_data).fit(method="qr")
+        except ValueError:
+            continue
         samples.append(mod.params)
     return samples