Apart from the problem you are writing about I'd like to ask you to explain how you interpret counted frac_MCV - for me it has no sense at all to summarize most_common_freqs.
Please rethink it and explain what was the idea of such SUM ? I understand that it can be some measure for ratio of NULL values but only in some cases when n_distinct is small.
regards
Statistics: n_distinct, MCV, histogram
Useful to check statistics leading to bad join plan. SELECT (SELECT sum(x) FROM unnest(most_common_freqs) x) frac_MCV, tablename, attname, inherited, null_frac, n_distinct, array_length(most_common_vals,1) n_mcv, array_length(histogram_bounds,1) n_hist, correlation FROM pg_stats WHERE attname='...' AND tablename='...' ORDER BY 1 DESC;
