如何解读chi-square 检验结果?

救助:如何解读chi-square 检验结果?

我想请教许博士和其他会员, 我用Antconc 分别算出了三个平行语料库中的4类从句的次数,想比较三个语料库之间每一类从句分布次数的差异。我想用chi-square 来验证差异的显著性。得出了以下结果:
Chi-Square Tests   Value      df      Asymp. Sig. (2-sided)

Pearson Chi-Square   25.736a    6         .000
Likelihood Ratio     26.114    6         .000
Linear-by-Linear Association 7.857  1         .005
N of Valid Cases 942

a. 3 cells (25.0%) have expected count less than 5. The minimum expected count is .90.

请问这个结果如何能反映出库与库之间每一类从句次数的差异?
是否我的数据输入有问题?
 
Last edited:
回复: 如何解读chi-square 检验结果?

The chi-squared values given appear to suggest the differences are significant, but the chi-square tests are unreliable here becuase of low frequencies, as indicated in the note: 3 cells (25.0%) have expected count less than 5. You may need to use Fisher's Exact Test (available in SPSS) to test for the exact significance level.
 
回复: 如何解读chi-square 检验结果?

Many thanks for your prompt reply. Does it mean that chi-square is not suitable for a significant test in my study? Is it ok to solely compare the percentage of each type of clauses in each corpus? I found that some corpora studies compare percentages rather than running any statistical tests. I wonder whether this practice is acceptable.
 
回复: 如何解读chi-square 检验结果?

Chi-squared test is unsuitable in this case, unfortunately.

You can simply compare proportional data, and many people do this, but proportions are merely descriptive statistics but not inferential statistics.


Many thanks for your prompt reply. Does it mean that chi-square is not suitable for a significant test in my study? Is it ok to solely compare the percentage of each type of clauses in each corpus? I found that some corpora studies compare percentages rather than running any statistical tests. I wonder whether this practice is acceptable.
 
回复: 如何解读chi-square 检验结果?

Thanks a lot. So it seems that an inferential test is indeed necessary to test statistical significance. But how to run a fisher's exact test? Could you provide some guidance?
By the way, I bought the newly published book of Using Corpora: A Practical Coursebook. It is really practical and theoretical. Good for reading.
 
回复: 如何解读chi-square 检验结果?

Reject alternate, Your p value has to be 0.05 or lower to accept alternate hypothesis, (or to reject the null hypothesis) though you can mention that the value approaches statistical significance.
 
Back
顶部