具有线性回归特征的多组实验结果对比分析方法的研究

研究试验参数变化情况下．具有线性回归特征的多组试验结果差异性的对比分析方法。数学上谈问题转化为对回归直线的常数口和比例系数6两组随机样本的F检验。当这两个指标均无显著差异时，认为两组(或多组)试验数据无显著差异。谈法将判断的范围外推到了整个实数域，因此要求条件过高。本文提出了重合度法，即用试验参数的共同区域内回归直线置信带的面积重合度．判断实验结果的一致性。对试验数据的分析表明，该方法具有明显的合理性．但其临界值还需进一步研究确定。

An Approach for Testing Difference among Data Characterized with Linear Regression

Approaches for testing the differences between/among the experimental data of two or more groups are studied under variable test parameters and can be linearly regressed. One approach, from mathematical point of view, is to translate the test into comparson of the constant a and proportion b of linear regression by using F test. When there are no differences for both F tests, experimental data are considered as no difference. This approach extends the estimation into all real number fields, so the requirement for passing the test is too strict. A new approach, superposition rate of confidence bands (SRCB) for linear regression of the experimental data in common area, is proposed. Analysis for several group data shows that the SRCB is obvious rational, but further study is needed to define the critical value r 0 mathematically.