: --
:
※ 發信站: 批踢踢實業坊(ptt.cc)
: ◆ From: 120.126.38.177
: → yhliu:因為這些 sub-tables 的邊際分布不同, 若不做控制(分層), 02/02 17:25
: → yhliu:合併表的 marginal odds-ratio 會與 conditional odds-ratio 02/02 17:26
: → yhliu:不同. 就像 y=a+b1*x1+b2*x2+e 的 b1 與 y=a'+b'*x1+e' 的 02/02 17:27
: → yhliu:b' 會不同. 02/02 17:28
很感謝您的回答! 我好像有點懂了..(雖然沒想到實際的例子)
另外請教個問題，這篇文章
http://udel.edu/~mcdonald/statcmh.html
他說如果只是要做 hypothesis testing 則 C-M-H 不需要要求各
sub-table odd-ratio 相同，如果需要推估 odd ratio 值則需要，這個
說法您覺得可以認同嗎?
節錄兩段於下:
Some statisticians recommend that you test the homogeneity of the
odds ratios in the different repeats, and if different repeats
show significantly different odds ratios, you shouldn't do the
Cochran–Mantel–Haenszel test. [略]
Other statisticians will tell you that it's perfectly okay to use
the Cochran–Mantel–Haenszel test when the odds ratios are
significantly heterogeneous. The different recommendations depend
on what your goal is. If your main goal is hypothesis testing—
you want to know whether legwarmers reduce pain, in our example—
then the Cochran–Mantel–Haenszel test is perfectly appropriate.
A significant result will tell you that yes, the proportion of
people feeling ankle pain does depend on whether or not they're
wearing legwarmers. If your main goal is estimation—you want to
estimate how well legwarmers work and come up with a number like
"people with ankle arthritis are 50% less likely to feel pain if
they wear fluorescent pink polyester knit egwarmers"—then it
would be inappropriate to combine the data using the Cochran–
Mantel–Haenszel test. If legwarmers reduce pain by 70% in the
winter, 50% in the spring, and 30% in the summer, it would be
misleading to say that they reduce pain by 50%; instead, it would
be better to say that they reduce pain, but the amount of pain
reduction depends on the time of year.
我目前想要做得不是要去推估 odd ratio，只是想知道 X, Y 有沒有相關
所以依照這個講法應該可以不管 odd ratio 直接使用 C-M-H。
另外還有一個問題，我覺得我的問題應該是 two-tail，我要測約兩萬筆
資料(基因)，對某些基因來講，可能是某個方向，其他可能是別的方向，
，但 two-tail C-M-H 是 significant 的時候，我需要知道他是哪個方向，
但我又不要去計算它統一的 odd ratio 時，我要怎麼判斷呢?
--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 120.126.38.177
※ 編輯: huggie 來自: 120.126.38.177 (02/03 14:12)
1F：→ bmka:hypothesis testing 跟estimation 不是同一件事 02/03 20:58
2F：→ yhliu:從 C-M-H 統計量的公式來說, 如果各 sub-table 的關聯方向 02/03 23:05
3F：→ yhliu:不一, 或有許多 sub-tables 的關聯太小, 則此檢定效力受影響 02/03 23:09
4F：→ yhliu:尤其是前一種情形, 正負向關聯會相互抵消, 使重要的條件關聯 02/03 23:10
5F：→ yhliu:無法檢測出來. 02/03 23:11
6F：→ yhliu:Conditional independence against full model 的檢定可用 02/03 23:11
7F：→ yhliu:sum of chi-squares 或其他加總型的統計量. 02/03 23:12