User Tools

Site Tools



This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
hcistats:ttest_jp [2017/03/08 01:47]
Koji Yatani [Effect size for an unpaired t test]
hcistats:ttest_jp [2017/03/08 01:52] (current)
Koji Yatani [A paired t test]
Line 65: Line 65:
 ---- ----
-=====A paired ​test===== +=====対応のあるt検定===== 
-You should use a paired ​test if you do a within-subject design. What a paired ​test does is to take differences between data in the two groups, and see whether the distribution of the differences is too different from the distribution. Because it uses the differences between the groups, ​**a paired ​test does not assume the variances of the population of the two groups are equal**. But it still assumes the normality. The null hypothesis is there is no significant difference in the means between the two groups. If the p value is less than 0.05, you reject the null hypothesis, and say that you find a significant difference.+対応のあるt検定は被験者内要因があるときに使います.対応のあるt検定がやっていることをおおざっぱに言うと,2つのグループの差を取って,その差の分布がt分布とどれくらい違っているかを計算しています.この「差を取る」という作業があるため,**対応のあるt検定では2つのグループの母集団の分散が同じであるという仮定は必要としません.**しかし,正規性は必要です.帰無仮説として2つのグループに差がない(つまり,差の平均が0である)こととして,検定を行います.
 ---- ----
hcistats/ttest_jp.txt · Last modified: 2017/03/08 01:52 by Koji Yatani