(32(2):122-128)Studies on Sampling Method of Rice Agronomic Characters－I. Comparison between Crops

本文探討在不同期作下，試區取樣時水稻重要農藝性狀之適當取樣單位及樣品大小，並作產量構成要素與實粒重關係之比較。以臺農67 號為材料，分70 年二期作與71 年一期作調查株高、實粒重及產量構成要素等12 個性狀，並作○1取樣單位間與取樣單位內變異性之比較，○2取樣單位大小之影響，○3取樣大小之影響及○4實粒重與其他性狀之相關。結果顯示一、二期作各性狀取樣單位內之變異較取樣單位間之變異為大，且取樣單位間之差異達極顯著，表示試區內之植株間仍有某種程度之變異。除產量外，相鄰四株（2 株× 2 株）之取樣單位可應用於一期作，而相鄰四株（2 株× 2 株）或相鄰六株（2 株× 3 株）之取樣單位較適於二期作。樣品大小在一期作時每試區內3 個取樣單位已足夠，二期作時每試區內4 個取樣單位較準確。試區內之偏性因素除應盡力控制外，尚應盡可能在試區內逢機取樣，以使可能導入試驗結果之偏性因素能有機會互相抵消。實粒重與穗數之相關極高（一期作0.74 ，二期作0.81 ）但與每穗穎花數，結實率及千粒重等「導出性狀」（derived character ）之相關很低，因此作選拔或作產量預測時應以4 個產量構成要素之乘積作指標，而不能以前述導出性狀為依據。 This paper investigated the appropriate sampling unit and sample size, under plot sampling, for various agronomic characters of rice. Using Tainung 67 as experimental variety, 12 agronomic characters including grain weight and yield components were measured for the second crop of 1981 and for the first crop of 1982. Analyses performed were (1) variability comparison among and within the sampling units, (2) effects of the size of sampling unit, (3) effects of the smple size and (4) the relationship between yield components and the weight of filled grains. Data revealed that variation within individual sampling units was always larger than that among different sampling units. Furthermore, the difference between sampling units ware also highly significant for both crops and all measured characters, indicating that some sort of variation existed among plants within the same plot. Excepting yield, 4 adjacent hills (2-hill x 2-hill) should be counted as a sampling unit for the first crop, and either 4 adjacent hills (2-hill x 2-hill) or 6 adjacent hills (2-hill x 3-hill) should be used for the second crop. A sample size of 3 sampling units for the first corp and 4 sampling units for the second crop are considered appropriate. The selection of sampling units should be random whenever possible in order to cancel out any potential bias associated with the experimental results. The weight of filled grains correlated closely with the number of panicles per hill (0.74 for the first and 0.81 for the second crop). However, there was very little correlation with such “derived” yield components as the number of spikelets per panicle, % of filled grains, and the 1000-grain weight. Therefore, the product total of the four yield components, rather than any derived one should be used as an index in making selections or predicting yield