I am trying to analyze a paired dataset where each subject had a test performed different ways from the same sample. The test gives an answer that is five ordinal levels (Very High, High, Medium, Low, Very Low).
The data looks something like this in R
Data= read.table(text="
1 VeryHigh VeryHigh
2 VeryHigh VeryHigh
3 VeryHigh VeryHigh
4 VeryHigh Medium
5 VeryHigh Low
6 VeryHigh Low
7 High VeryHigh
8 High VeryHigh
9 High High
10 High High
11 High High
12 High High
13 High Medium
14 High Medium
15 High Medium
16 High Medium
17 High Low
18 High Low
19 High Low
20 Medium Medium
21 Medium Medium
22 Medium Medium
23 Low VeryHigh
24 Low Low
25 Low Low
26 VeryLow VeryLow
27 VeryLow VeryLow
28 VeryLow VeryLow
29 VeryLow VeryLow
30 VeryLow VeryLow
31 VeryLow VeryLow
32 VeryLow Low
33 VeryLow Low
")
colnames(Data) <- c("StudyID", "Method1", "Method2")
Data$Method1 <- factor(Data$Method1, levels=c("Very High", "High", "Medium", "Low", "Very Low"))
Data$Method2 <- factor(Data$Method2, levels=c("Very High", "High", "Medium", "Low", "Very Low"))
table(Data$Method1, Data$Method2)
What I am trying to do is figure out is whether the two methods are similar or if one is reading a bit higher (or lower) than the other.
Because that focuses on the off diagonal, I thought the appropriate test would be Cochran's Q but I believe that is only for a binary response and mine actually has five levels.
Is there an alternative analysis for this data?
