甘藷田間試驗取樣技術之研究
Other Title
Studies on the Sampling Techniques in Field Trials of Sweet Potato
Journal
中華農學會報
Journal Issue
新73
Pages
9-22
Date Issued
1971-03
Author(s)
李良
Liang Li
Abstract
本研究係利用4個甘藷均度試驗，兩個為兩期稻田冬作及兩個為單期稻田冬作所得之資料，探討本省甘藷主要地區，在不同環境下，田間試驗的適用試區及合理而最經濟高效的規劃等，以期使甘藷試驗研究工作，有合理試驗技術之理據，減少試驗所費之勞資及增進試驗結果之準確性。
根據常態性測驗結果，本省兩期稻田冬作甘藷田間試驗，彰化地區，最小取樣單位，不宜小於行長6公尺以上之單行區。新竹地區，不宜小於行長8公尺以上之單行區。單期稻田冬作之甘藷田間試驗，雲林及臺南兩個地區，最小取樣單位，不宜小於行長6公尺以上之單行區。
應用最大曲度及SMITH氏變方律兩種方法，研究適合之試區大小，就試驗之準確性及實用上等之條件而論，兩期稻田冬作，彰化區最適之試區為6~12公尺之3行區，適合試區面積在18~36平方公尺之間。新竹區適合試區行長為8~12公尺之3行區，面積在24~36平方公尺之間。單期稻田冬作，雲林區最適試區之行長為6~12公尺之2行區，試區面積為12~24平方公尺。臺南區適當之試區為行長6~12公尺之2~3行區，試區面積在12~36平方公尺之間。
由等面積下之比較均方比F值顯著性測驗結果，得知同面積各種不同形狀試區間之F值，均無顯著之差異。然在田間試驗規劃上，因試區形狀與試地土壤肥力變異情形有相當的關係，故在試驗地之變異性大時，試區之形狀，實值得重視。
用土異係數b值來決定田間試驗之區集大小，很有採用之價值。若試驗地b值高，因每區集內所含試區數多少對相對效率之增加不大，故應採用完全區集設計為宜。b值低時，則縮小區集面積，可提高相對效率，故宜採用不完全區集設計為優。
適合之區集形狀係隨土壤肥力差異存在情況而定，故試驗地土壤肥力之變異方向及其分布情形等，為決定區集形狀之一重要因子。
適當的重複次數，係視試地之變異性大小；試項間之平均差數百分率大小；試項數多少；試驗設計方法及所需顯著性測驗水準高低，及獲得與零有顯著差數之最高適當機率等條件之變動而定。當上述之各項條件均考慮時，則試驗所需最少的重複次數，可應用r≥2(σ/δ) ^2(t1+t2)^2式估算。
Four uniformity trials, two late fall plantings and two winter relay-interplantings were conducted at four different locations in 1967-68 to obtain information on plot sizes and shapes, block sizes and shapes, number of replicates and cost relationships for conducting yield tests of sweet potato.
The results from the test of normality indicated that the smallest sampling unit in sweet potato experiments of relay-interplanting should be not than 6 m long in single row plot in Changhua and the plot of 8 m long in single row in Hsinchu. In the late fall planting, the smallest sampling unit should be at least the one single row plot of 6 m long in Yunlin and Tainan areas.
Smith's regression method and maximum curvature technique were used to estimate the optimum plot size. From all considerations evaluated, the optimum plot size for the relay-interplanting was estimated to be 6-12 m long and 3 rows wide with an area of 18-36 ㎡ in changhua and 24-36㎡ , or 3 rows wide by 8-12 m long in Hsinchu. In the late fall plating, the optimum plot size was considered to be 6-12 m long and 2 rows wide with an area of 12-24㎡ in Yunlin and 12-36㎡ , or 2-3 rows wide by 6-12m long in Tainan.
The F-test indicated that the comparable variances for differently shaped plots composed of equal numbers of basic units did not differ significantly for any of the comparisons. However, in the planning of field experiments, the shape of plot was directly related to the pattern of variability in the experimental land. Therefore, plot shape may be of considerably greater importance for experiments conducted on areas that are known to be highly variable.
By use of the coefficient of soil heterogeneity in deciding the efficient size of block is most desirable in planning of field experiments. If the field to show high values of b, the relative efficiency of blocks with few plots compared with blocks of many plots is not high and the simpler designs should be used. For low values of b, the use of layouts designed to reduce block size becomes more important.
The best shape of block or replication to be used is dependent upon the presence of fertility gradients. Therefore, the direction of soil fertility gradients and their distributing manner on the experimental site is one of important factor in determining the shape of block.
The suitable number of replication depends upon: (1) the true coefficient of variability, (2) the true difference between treatments to be detected in percent of the mean, (3) the number of treatments, (4) the design of experiment, and (5) the level of significance and the probability of detecting a difference. When these factors are considered, the optimum number of replication for any given experiment was computed using the following formula: r≤2(σ/δ)^2(t1+t2)^2.
Subjects
甘藷
Publisher
中華農學會
Type
journal article