n**m
发帖数: 156 | 1 有8个group，我知道里面的人口比例
Asian White Hispanic Black Other
1 5% 30% 30% 30% 5%
其他7个group也是类似的数据
其中有4个group是treatment,四个group是control，要测treatment vs control人口比
例是不是一致 |
w*******9
发帖数: 1433 | 2 抛砖引玉：如果不想group成2个组分析的话，可以考虑一个split plot design的分析
方法。use treatment as whole plot and race as split plot while the group
number as the (random) block effect. |
n**m
发帖数: 156 | 3 有8个group，我知道里面的人口比例
Asian White Hispanic Black Other
1 5% 30% 30% 30% 5%
其他7个group也是类似的数据
其中有4个group是treatment,四个group是control，要测treatment vs control人口比
例是不是一致 |
w*******9
发帖数: 1433 | 4 抛砖引玉：如果不想group成2个组分析的话，可以考虑一个split plot design的分析
方法。use treatment as whole plot and race as split plot while the group
number as the (random) block effect. |
w*******n
发帖数: 469 | 5 Chisquare? use the percentage as weight? |
n**m
发帖数: 156 | 6 Chi-square is test highly sensitive to the count. I think only independent
frequency data is good for the test. Very likely you will get a significant
statistic if using percentage.
【在 w*******n 的大作中提到】
: Chisquare? use the percentage as weight?
|
