在試驗設計上，有時為了試驗目的，或者為了試驗上的方便而採用較為特殊的設計法，如隨機完全區集設計(Randomized Complete Block Design：RCBD)，是農業試驗上常採用的設計法。而由試驗所得的資料,如果資料中，各處理因子之間的處理組合，其各處理組合中重複數不等或其中有處理組合缺失的情況下，冒然的採用一般的變方分析法會影響試驗結果的可靠性。
本研究主要的目的是在利用均值模式(Means Model)來分析各處理因子之間，各處理組合中重複數不等或有處理組合缺失的情況下之綜合變方分析法，而後以此結果對實際的田間資料進行分析。本試驗研究資料為毛豆品種比較試驗結果記錄(青株重公頃產量)。各個參試品種在各個年度及各個試場下，因發芽率不佳或受到環境的影響而發生品種產量缺失或(及)各個品種產量間觀測值重複數不等的情況，本文利用均值模式(Means Model)來探討參試品種在各個年度及各個試場下產量分析效果。

Sometimes for the reason of experimental purposes or experimental conveniences, we will use special design methods. For instance, randomized complete block design is usually used in agricultural experiments. If the data arising from these experimental designs with unequal numbers in each treatment combination, or some treatment combinations are missing, it appears to be less reliable of traditional analysis of variance.
This study is on the basis of the means model, which is a combined analysis of variance, to analyze the unequal numbers in each treatment combination or missing treatment combinations mentioned above. Practical field data are used as an example to illustrate the methods. The data of this experiment is the fresh weight per hectare of six varieties of green-soybean. Missing yields or(and) unequal numbers in each variety may exist, which is due to the insufficient germination rates or the effect of environment. This study is to use means model to probe into the yields of these six varieties of green-soybean in each year and location.

中文摘要 .................................................I
英文摘要 ................................................II
一、前言 .................................................1
二、文獻探討 .............................................3
三、不完全資料變方分析之誘導 .............................6
四、不完全資料綜合變方分析法 ............................29
五、試驗資料變方分析之前提檢定 ..........................37
六、實測資料統計分析結果 ................................43
七、結論 ................................................57
八、參考文獻 ........................................... 59

中文參考資料
1. 沈明來（1999）試驗設計學。台北，九州圖書文物有限公司。
2. 高月庭（1995）無前題之綜合變方分析。台北，國立台灣大學農藝學
研究所生物統計學組碩士論文。
3. Bancrof, T, A. (1964) Analysis and inference for
incompletely specified models involving the use of
preliminary test(s) of significance. Biometrics20：427-442.
4. Bancroft, T, A.(1968) Topics in Intermediate Statistical
Methods. Ames, Iowa, U.S.A.
5. Cochran, W. G. (1951) Testing a linear relation among
variances. Biometrics 7：17-32.
6. Kempthorne, O. (1952) The Design and Analysis of
Experiments. New York：John Wiley & Sons.
Milliken, G. A. and Johnson, D. E. (1989) Analysis of Messy
Data (Volumn Ⅱ：Nonreplicated Experiments). New York：Van
Nostrand Reinhold.
7. Milliken, G. A. and Johnson, D. E. (1992) Analysis of Messy
Data (Volumn Ⅰ、Ⅱ：Designed Experiment). ：New York
：Chapman &Hall.
8. Sendecor, G. W., and Cox, G. M. (1935) Disproportionate
subclass numbers in tables of multiple classification. Ames,
Iowa, U.S.A.
9. Wilk, M. B., and Kempthorne, O. (1955) Fixed, mixed and
random models. Fournal of the American Statistical
Association 50：1144-1167.
10. Yates, F. (1934) The analysis of multiple classifications
with unequal numbers in the different classes. Fournal of
the American Statistical Association 23：108.