0

Here I described my general situation. How to calculate 95%CI between means in R?

Community
  • 1
abc
  • 1,747
  • 3
  • 17
  • 32
  • You may want to reiterate the details of your design on this question. Is your design repeated measures or between subjects? Are you going to assume normally distributed errors? – Jeromy Anglim Sep 26 '11 at 00:40
  • @Jeromy Anglim: thanks for your questions :)! [Here](http://stats.stackexchange.com/questions/11887/is-this-design-a-one-way-repeated-measures-anova-or-not) I was answered that my design is one way RM ANOVA. And it is within-subject as far as I understand... And I want to calculate 95%CI instead of post hoc tests... – abc Sep 26 '11 at 03:11
  • ...as [here](http://stats.stackexchange.com/questions/15361/what-type-of-error-correction-do-i-have-to-use-after-wilcoxon-matched-pairs-test) @John advised me to *"report the standard confidence intervals of the effects in question"*. Jeromy, I have a counter-question: the term *within (and/or between)-subject design* is more general than, for instance, ANOVA or RM ANOVA and so on, isn't it? – abc Sep 26 '11 at 03:11
  • I guess I'm trying to set the question title and the text of this question to stand alone for future readers. Thus, I think the title of this question and the content should make it clear what kind of confidence interval you are interested in. The two most common confidence interval scenarios for comparing means are probably those for repeated measures t-test and independent groups t-test, hence my question. – Jeromy Anglim Sep 26 '11 at 03:31
  • @Jeromy: please suspend the question. I have to think about the modification... – abc Sep 26 '11 at 04:05
  • We don't have an actual "suspend" mechanism, but we can close a question and you can later flag it for reopening. – whuber Sep 26 '11 at 13:48

1 Answers1

2

That depends on a bunch of things. One easy way to do it might be just to calculate t-tests for each one. Part of the output is the confidence interval of the effect.

t1 <- t.test( A, B, paired = TRUE )
t1$estimate # the effect in question
t1$conf.int # the confidence interval of the effect
John
  • 21,167
  • 9
  • 48
  • 84