`ValueError` when running on discovery

[TheChymera]

As reported by Austin we ran into another error on discovery (stderr, stdout).

The crux of it seems to be that the ANOVA is failing:

----  Messages for py:session9:default  ----
  Traceback (most recent call last):
* PythonTeX stderr - error on line 17 in "results.tex":
    File "<outputdir>/py_session9_default.py", line 429, in <module>
      print(pytex.formatter(boilerplate.anova(factor='C(Q("PA rel. Bregma [mm]"))')))
    File "/opt/src/opfvta/lib/boilerplate.py", line 143, in anova
      summary = sm.stats.anova_lm(ols, typ=typ, robust='hc3')
    File "/usr/lib/python3.10/site-packages/statsmodels/stats/anova.py", line 349, in anova_lm
      return anova_single(model, **kwargs)
    File "/usr/lib/python3.10/site-packages/statsmodels/stats/anova.py", line 83, in anova_single
      return anova3_lm_single(model, design_info, n_rows, test, pr_test,
    File "/usr/lib/python3.10/site-packages/statsmodels/stats/anova.py", line 252, in anova3_lm_single
      f = model.f_test(L12, cov_p=cov)
    File "/usr/lib/python3.10/site-packages/statsmodels/base/model.py", line 1765, in f_test
      res = self.wald_test(r_matrix, cov_p=cov_p, invcov=invcov, use_f=True, scalar=True)
    File "/usr/lib/python3.10/site-packages/statsmodels/base/model.py", line 1841, in wald_test
      LC = DesignInfo(names).linear_constraint(r_matrix)
    File "/usr/lib/python3.10/site-packages/patsy/design_info.py", line 536, in linear_constraint
      return linear_constraint(constraint_likes, self.column_names)
    File "/usr/lib/python3.10/site-packages/patsy/constraint.py", line 419, in linear_constraint
      return LinearConstraint(variable_names, coefs)
    File "/usr/lib/python3.10/site-packages/patsy/constraint.py", line 62, in __init__
      raise ValueError("must have at least one row in constraint matrix")
  ValueError: must have at least one row in constraint matrix

It's reported via TeX, but this is in the end a Python error, caused by either:

1. Missing data on account of quiet preprocessing errors
2. Updates in the statsmodels/patsy packages.

Since we pin the versions, I'm pretty sure it's (1), but I'll need to meet with @asmacdo again and check.

[TheChymera]

@asmacdo if you could check the lengths of our stats summary files on discovery:

[dark]~/docsrc/opfvta ❱ for i in data/*csv; do wc -l ${i}; done
18 data/features_structural.csv
79 data/functional_significance.csv
79 data/functional_t.csv
34 data/groups.csv
10 data/implant_coordinates_block.csv
10 data/implant_coordinates.csv
10 data/implant_coordinates_phasic.csv

That would give us a good indicator whether data was dropped on discovery. If your numbers are larger, that just means my last local run dropped some data.

[asmacdo]

This was fixed way back