Inter-rater reliability

[jykim]

Hello,

I have 50 survey questions: q1 - q50
I would like to perform inter-rater reliability between two raters for each of these questions using

 kap varname1 varname2 if in weight , options Weights

The problem I am facing now is that I need to reshape the data first because the data recorded the observations like the below.

ID	Q1	Q2	Q3
11	1	1	1	patient1 recorded by rater1
12	1	1	2	patient1 recorded by rater2
21	1	1	2	patient2 recorded by rater 1
22	1	1	1	patient 2 recorded by rater1

Could you please advise me how I can reshape this data to the format (question number.rater number) so that I can perform the test

 kap q1.1 q1.2

ID	Q1.1	Q1.2	Q2.1	Q2.2	Q3.1	Q3.2
1	1	1	1	1	1	2
2	1	1	1	1	2	1

Or is there any way that I can test inter-reliability test without reshaping?


Thank you so much for your comments.
I really appreciate your help.

[daniel klein]

You need to show us an excerpt of your real data. Is the ID a string or numeric variable? For reshaping you need one variable that identifies the subjects and one that identifies the raters. How to create such a variable depends on the details of your data.

Best
Daniel

[jykim]

Thank you so much for your feedback.

Here is the excerpt.
I just reduced the variables up to 5 variables. but there are 50 variables (questions).

uniquerecordnumber	record	rater	q1_colonoscopy	q2_antibiotics	q3_admission date	q4_infection	q5_status
11	1	1	Yes	Yes	19-Jan-16	Yes	Stable
12	1	2	Yes	Yes	19-Jan-16	Yes	Stable
21	2	1	Yes	Yes	15-Feb-16	Yes	Stable
22	2	2	Yes	Yes	15-Feb-16	Yes	Stable
31	3	1	Yes	Yes	1-Mar-16	Yes	Stable
32	3	2	Yes	Yes	1-Mar-16	No	Stable

Thank you so much for your help Daniel!

[daniel klein]

Better to use dataex (from SSC, also see the Statalist FAQs) for posting example data. Anyway, you want

drop uniquerecordnumber
unab vars : q*
reshape wide `vars' , i(record) j(rater)

If your variables are strings (which would have been obvious had you used dataex) you want to encode them before calculating inter-rater agreement.

Also see kappaetc (SSC) for alternative measures to the kappa coefficient.

Best
Daniel

[jykim]

Thank you so much, Daniel.
I will use dataex in the future.