Inter and Intra observer variability

[LarsFolkestad]

Dear Statalisters

I have hiered help to run some measurements for me. Before letting them start on all the data I want to test their inter- and intra observer variations.
They are all doing 10 measurements twice, to test the above.

This is a mock-dataset.

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(id measurments observer time)
 1   -.3124319 1 1
 1   -.2761542 1 2
 1   -.8676338 2 1
 1     .829529 2 2
 1   1.6160905 3 1
 1  -1.1825305 3 2
 2  -.14859566 1 1
 2    -1.32869 1 2
 2   .05580055 2 1
 2   1.8104898 2 2
 2  -1.5255077 3 1
 2  -1.5434698 3 2
 3   -.5675358 1 1
 3  -1.6896724 1 2
 3   1.3391538 2 1
 3    1.294992 2 2
 3   -.7828341 3 1
 3   -.9400627 3 2
 4   -.6450798 1 1
 4   .23616154 1 2
 4    .9733304 2 1
 4 -.001522023 2 2
 4    .5158845 3 1
 4   1.0015099 3 2
 5    -1.80597 1 1
 5    .7093297 1 2
 5   1.2016693 2 1
 5   1.1414266 2 2
 5   2.1341999 3 1
 5   -.4797522 3 2
 6   1.8456104 1 1
 6    .4957038 1 2
 6  -.04035825 2 1
 6   .16795605 2 2
 6  -.06805837 3 1
 6  -1.1547463 3 2
 7     .378075 1 1
 7   1.5130014 1 2
 7    .5927394 2 1
 7  -2.3594098 2 2
 7   1.0259678 3 1
 7   -.2045485 3 2
 8  -.10719195 1 1
 8    .3469145 1 2
 8     2.51908 2 1
 8    .8936414 2 2
 8   .07947483 3 1
 8   .04911203 3 2
 9    .3044299 1 1
 9   1.6418105 1 2
 9  -1.6078342 2 1
 9     .460919 2 2
 9   -.5155021 3 1
 9  -1.2407768 3 2
10  -2.6493924 1 1
10    -.308823 1 2
10    .7012202 2 1
10  -.17781775 2 2
10    .9778325 3 1
10   .14720653 3 2
end

where observer is the observer id (1,2,3) time is the first or second time the measurement was done.


now I want to calculate the inter-observer variation but how to do that best?
And the intra-observer variation form time 1 to time 2, same question?

Now further more, how to reshape the data so that I get the analysis that I want?

Thanks for taking the time.

Lars

[Joe Canner]

Lars,

What you probably want is the intra-class correlation coefficient, which is implemented by the Stata command icc. Possible usages might include:

icc measurements id
bysort time: icc measurements id observer 
bysort observer: icc measurements id time

There is no need to reshape the data to use this command.

Regards,
Joe

[LarsFolkestad]

Thank you. That made everything easy
Lars