多次調查減緩種子發芽試驗中不足變異之評估

The Assessment of Multiple-investigation on Reducing Under-dispersion in Germination Tests

陳思婷(Si-Ting Chen) ； 許鐈云(Ciao-Yun Syu) ； 郭育妏(Yu-Wen Kuo) ； 郭寶錚(Bo-Jein Kuo)

《作物、環境與生物資訊》 15卷3期 (2018/09) Pp. 159-168

https://doi.org/10.30061/CEB.201809_15(3).0002

摘要

種子發芽試驗(seed germination test)為種子檢查中相當重要的項目之一，國際種子檢查協會(International Seed Testing Association, ISTA)詳細規範發芽試驗流程，其中容許度(tolerance)可用來決定發芽試驗結果是否可靠，若試驗結果超出對應之容許度，該發芽試驗則須重新試驗(retest)，然而在ISTA之容許度中並未考慮不足變異(under-dispersion)。不足變異為重複間的觀測變異(empirical variance)小於二項取樣變異(binomial sampling variance)，若發芽試驗中不足變異存在，使用ISTA之容許度判斷發芽試驗是否須重新試驗，可能會得到錯誤之結論。目前有許多研究已確認發芽試驗中存有不足變異，且認為不足變異與試驗人員無意識地調整試驗結果有關。本試驗調查種子檢查室(Seed Testing Laboratory, Council of Agriculture)發芽試驗之歷史資料，發現種子來源、調查總次數與試驗期限為影響不足變異程度之主要因子，因此推測在發芽試驗中若設置多次調查，或許可減緩不足變異。藉由設計多次調查試驗，確認多次調查可藉由提高每次調查與每個重複結果之獨立性來減緩不足變異。未來，可根據本試驗之結果修改發芽試驗流程，避免試驗人員無意識地調整試驗結果並減緩不足變異，以得到更加可靠之發芽試驗結果。

關鍵字

種子發芽試驗 ； 不足變異 ； 多次調查 ； 觀測變異 ； 二項取樣變異

並列摘要

The seed germination test is one of the important item in seed testing. The International Seed Testing Association (ISTA) regulates the process of a germination test, and the tolerance is used to determine whether the result of a germination test is reliable. If the result of a germination test is out of the corresponding tolerance, the germination test should be retested. By definition, if the empirical variance between replicates is less than binomial sampling variance, it was called under-dispersion. However, the tolerance in the ISTA regulation does not consider the under-dispersion. When under-dispersion exists in germination tests, it may result in a wrong conclusion. Some studies have confirmed that the under-dispersion existed in germination tests because analysts unconsciously adjusted the result of germination tests. In this study, the data of historical germination tests in Seed Testing Laboratory, Council of Agriculture were collected. In the data, the seed source, frequency of counting per germination test, and duration of the test were factors which would affect the magnitude of under-dispersion. Therefore, we proposed a multiple-investigation in the germination test to reduce the under-dispersion. By designing the multiple-investigation experiment, it was confirmed that the multiple-investigation could reduce the under-dispersion by increasing the independence between each investigation and replicate. To avoid analysts unconsciously adjusting the result of germination tests and to reduce the under-dispersion, the process of a germination test can be modified based on our findings.

並列關鍵字

Seed germination test ； Under-dispersion ； Multiple-investigation ； Empirical variance ； Binomial sampling variance