==== 维尔克松T检验(The Wilcoxon T test)==== 对应于相关样本t检验,用于检验重复方差设计的两种处理条件之间的差异。 Corresponding to the correlation sample t test, and is used to test the difference between the two treatment conditions of the repeated variance design ---- === 维尔克松T检验的步骤 (Steps of Wilcoxon T test)=== == 1、确定虚无假设和备择假设 (Determine null hypothesis and alternative hypothesis)== H0:不同处理间没有显著差异。(There is no significant difference among different treatments.) H1:不同处理间有显著差异。(There are significant differences among different treatments.) == 2、计算差异分数并排序 (Calculate the difference scores and rank them)== 计算每个样本不同处理下的分数差,此处差异分数的排序是按照差异分数绝对值的大小进行排序。 Calculate the score difference under different treatments for each sample, where the difference scores are sorted according to the magnitude of the absolute value of the difference scores. == 3、计算正负差异分数的秩和 (Calculate the rank sum of positive and negative difference scores)== 排序后每个差异分数对应一个秩,分别计算正的差异分数的秩之和∑R+和负的差异分数的秩之和∑R-。 After sorting, each difference score corresponds to a rank. Then calculate the rank sum of the positive difference score ∑R+ and the rank sum of the negative difference score ∑R- respectively. == 4、取T值并与临界T值比较 (Take the T-value and compare it with the critical T-value)== T=min(∑R+,∑R-) 查表可以得到临界值Tcrit。 The criticle value Tcrit can be obtained by looking up the table. 若T<Tcrit,则拒绝H0。 If T<Tcrit, then reject H0. ---- === 大样本条件下的维尔克松T检验 === 当n>25时,T分布近似服从正态分布。 该正态分布的均值为: {{ ::t均值.jpg?nolink&150 |}} 方差为: {{ ::t方差.jpg?nolink&200 |}} 由此可以将T分数标准化为z分数,查z分数表进行比较。 ---- === 相同的等级和0分数 === 在Wilcoxon 检验中,有两类相同的等级: 1、不同被试在不同处理条件下的差异分数相同; 2、同一个被试在不同处理条件下得到相同的分数,因此算出的差异分数为0。 **对于0分数,有两种做法:** 1、去掉差异分数为0的被试; 2、将0差异分数均匀地分配在正负两组中,即在计算正负差异分数的秩和时将0分数的秩平均分配于两组,这样将增大T值。