Study and species weights give opposite signs for adjacent low output pair excess over random order in 631 MASTREE+ records

We tested whether low plant output occurs in adjacent year pairs more often than a random order predicts. We used 631 complete seed, fruit and cone records of 20 to 69 years from MASTREE+ version 2. They cover 229 named taxa and 99 source studies. Low output is at or below the 25th percentile of each record. Ties are kept. We compare observed pairs with the exact expectation for a random order of the same values. With equal study weights, the difference is -1.02 percentage points of adjacent pairs. The 95% interval is -1.87 to -0.15 points. With equal species weights, it is +1.07 points. Its interval is -1.04 to +1.39 points. Intervals use 9,999 resamples of whole source studies. Two tropical forest sources supply 64.6% of species weight but only 2.0% of study weight. The species estimate changes to -0.27 points when random order is limited to calendar decades. It is -0.53 points with blocks shifted by five years. Blocking also changes the eligible pairs. Comparisons on the same pairs show that both choices can matter. The observed sign depends on the source mix and the reference time scale. These data do not establish one global pattern of consecutive poor years. Low output is relative to each record and does not measure food shortage. The data mix methods and spatial units. Sources can overlap. We preserve source data, provenance, a plan, code, results, two figures and independent checks. The contribution is a reproducible analysis of time order. It is not a test of climate causes or a new pair formula.

plant reproduction · MASTREE+ · seed output · time series · source dependence · masting · random order