0

I have data of the structure below. im confused about using paired sample t-testing vs ANOVA. Paired T-tests make sense to me, but im worried; because there are 7 outcome measures, am i violating some assumption with t-test, and thus should use a form of ANOVA? if so, which type, as I am very confused. Thanks for any help!

enter image description here

AmielMatt.
  • 11
  • 2
  • 1
    Can you elaborate on your data, what are these variables that you are showing us, what do they represent? Also some context on the design and experimentation. – user2974951 Apr 07 '20 at 05:29
  • Sure, sorry. So these are continuous numerical variables. Two time points, n=6 at each time point. At each time point 7 outcomes are measures for each subject. – AmielMatt. Apr 07 '20 at 05:36
  • Also these are matched. So for example where you see "Outcome1 - Subject 1 in the Time 1 group; thats the same subject being measured again on the same variable at Time 2. – AmielMatt. Apr 07 '20 at 05:38
  • See https://stats.stackexchange.com/questions/3466/best-practice-when-analysing-pre-post-treatment-control-designs. $n=6$ is low, maybe tell us much more about the practical context. What are (in real life terms) outcome1, ..., outcome6? – kjetil b halvorsen Apr 07 '20 at 13:06

1 Answers1

0

It depends on the dimension in which the outcome of the test will be useful for you. If you want to find whether there is a change in the values from time to time , irrespective of the outcome types, you can perform the paired t-test, for which you have the melt the data.

Let's say that your data looks like this.

enter image description here

head(data)

library(data.table)

set1<- data[c(1,2:7)]
set2<- data[c(1,8:13)]
mset1<-  melt(set1)
mset2<-  melt(set2)
colnames(mset1)<- c('Outcome', 'Subject', 'time1')
colnames(mset2)<- c('Outcome', 'Subject', 'time2')
mdata<- as.data.frame(cbind(mset1, mset2[3]))
model1<- t.test(mdata$time1, mdata$time2, paired = T)
model1

Output

Paired t-test
data:  mdata$time1 and mdata$time2
t = -1.4521, df = 41, p-value = 0.1541
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-13.433714   2.195619
sample estimates:
mean of the differences 
              -5.619048
Mohanasundaram
  • 646