Concord command syntax issue

[Frank Carrera]

Hello,

What is the syntax of the concord command in STATA 13.0 to graph the limits of agreement of the average percentage difference of two continuous measurements?
Thanks

[Nick Cox]

I don't think concord (Stata Journal, as you are asked to explain) supports "average percentage difference", whatever that is, as a measure of agreement.

[Frank Carrera]

Thank you Sr. I appreciate your response.
The Bland-Altman plot analysis is a simple way to evaluate a bias between the mean differences, and to estimate an agreement interval, within which 95% of the differences of the second method, compared to the first one, fall. Data can be analyzed both as unit differences plot (Figure A2 at the image below) and as percentage differences plot (Figure A3). This latter is what i need to do on STATA. I only know the syntax of the command to get a figure like A2: concord rater1 rater2, loa. Thanks

[Nick Cox]

concord doesn't support that.

The graphs shown mush together two kinds of comparison. If it makes sense to compare methods using differences then A2 is appropriate. If it makes sense to compare methods using ratios then look rather at log(A / B) = log A - log B compared with log(A B) / 2 = (log A + log B) / 2 or equivalently use geometric means. To do that, just take logs before feeding to concord.

A2 suggests to me pattern beyond an overall mean difference.

I get the impression that you're quoting from a published study but have loosely similar data. If so, posting the data might allow precise suggestions.

Please note the spelling Stata as explained at the end of the FAQ Advice.

18. What was that comment on Stata, not STATA, about?

The correct spelling is “Stata”, please, not “STATA”. Several of the most active experts on the list can get a little irritated if you get that wrong, although you are free to regard them as pedantic. More importantly, if you write “STATA” you are making it obvious that you didn't read this guide carefully and to the end.

[Frank Carrera]

If it makes sense to compare methods using ratios then look rather at log(A / B) = log A - log B compared with log(A B) / 2 = (log A + log B) / 2 or equivalently use geometric means. To do that, just take logs before feeding to concord.

Can please give a example of that? Thank you.
This are the data

method A	method B
1470	2705
1970	3138
1805	2670
1856	2975
1734	2665
1833	2788
1890	3033
1649	2488
1575	3000
1520	2440
2123	3050
2041	2700
1704	2393
1769	2970
2825	3320
1693	3000
1525	2320
1946	2720

[Nick Cox]

Something like this:

clear
input methodA    methodB
1470    2705
1970    3138
1805    2670
1856    2975
1734    2665
1833    2788
1890    3033
1649    2488
1575    3000
1520    2440
2123    3050
2041    2700
1704    2393
1769    2970
2825    3320
1693    3000
1525    2320
1946    2720
end

gen logA = log10(methodA)
gen logB = log10(methodB)

concord logA logB

But in in your case concord methodA methodB seems more appropriate. There is a massive systematic difference between methods.