Compare difference between recorders' procedures

[Joa14]

I have a dataset that different people recorded.  Each recorder must have used their own way of recording the variable CO, so I want to check whether there is a significant difference between the different recorders. This way, I can check if I can trust the data or not. The CO measurements were all taken at 31 different points (not duplicates). Any idea about a test and code I can use for that? Thanks a lot!

Values below. 

CO (%)	Recorder (Initial)
1.56	J
0.52	J
1.56	J
1.56	J
3.12	J
8.32	J
4.68	J
9.36	J
56.16	J
73.32	J
1.04	K
2.6	J
8.32	J
5.2	J
4.16	J
2.08	A
2.6	J
6.76	J
15.6	J
0.52	J
7.28	J
7.8	K
3.64	J
22.36	J
17.68	K
0.52	K
4.16	J
3.12	J
3.12	J
2.08	J
13.52	J

[Ksharp]

When I am looking at your question at first place, I think you should try ANOVA.
Like PROC GLM or PROC NPAR1WAY (nonparameter version of ANOVA).
But when I looked at your data, you only have 31 obs and Each recorder has different obs,
and you said "were all taken at 31 different points", I think that make question more complicated.
different points may have different mean of CO,you also should take into account of this effect ,
that lead you to make a MIXED model,
Like:
proc mixed data=have;
class recorder points;
model co= recorder ;
random int/subject=points;
lsmeans recorder/diff pdiff cl lines;
run;

And @StatDave @SteveDenham @lvm could give you better suggestion.

[SteveDenham · Accepted solution]

I think @Ksharp had it right the first time - PROC GLM or NPAR1WAY. No need for a mixed model here. There are 31 experimental units (the measurements) where each received one of three treatments (J, K or A). This code tests for differences between recorders using PROC GLM:

proc glm data=yourdata;
class recorder;
model value=recorder;
means recorder/lsd lines hovtest; /* you may want to substitute tukey for lsd */
lsmeans recorder/pdiff;
run;
quit;

And this for the nonparametric analysis:

proc npar1way data=yourdata;
   class recorder;
   var value;
run;

This gives several different approaches, but the easiest to consider is the Kruskal-Wallis test. The global test is not significant, so no post hoc pairwise analysis is done.

Now for the caveats: Of the 31 observations, 26 are for recorder J, 4 are for recorder K and 1 is for recorder A. This imbalance makes comparisons difficult, especially with recorder A. Variability is essentially set by recorder J.  None of the tests for difference are significant but this does NOT mean the recorders are equivalent. It may just mean that there is insufficient data to detect a real difference.

SteveDenham

[Joa14]

Thank you so much for this well-thought answer. I appreciate it!

[Joa14]

Thanks a lot @Ksharp!
PROC GLM was the solution I needed! Thanks!