Agreement Study

Luiz RR, Costa AJL, Kale PL, Werneck GL. Conformity assessment of a quantitative variable: a new graphical approach. J Clin Epidemiol 2003; 56:963-7. As noted in Liao [16], sample size is an increasing function of tolerance β but a decreasing function of the α discordance rate. The rate of discordance and the probability of tolerance play a similar role to that of the significant level and power in a hypothesis test parameter of the Neyman Pearson frame. More samples in k > 0 than k = 0 are needed to obtain the same probability of tolerance. For example, if k = 0 is n = 59, the conclusion of agreement with the discordance rate α = 0.05 can be quantified with a tolerance probability β = 0.95. However, if k = 1 (i.e. there is a pair of discrepancies) and with the sample size n = 59, the conclusion of compliance with the discordance rate α = 0.05 can be quantified with a tolerance probability β = 0.80. To have the same probability of tolerance β = 0.95, a large sample n = 93 is required in this case. Figure 3 clearly shows that all 32 flat-rate differences fall within the concordance interval limits at the α = 0.05 mismatch rate, i.e.

all 32 pairs are the concordance pairs. Thus, the concordance between the new test and the current test to measure the power of the rate of discordance α = 0.05 can be quantified with a probability of tolerance β = 0.80. A sound interpretation of a compliance analysis must explicitly demonstrate its dependence on clinical/scientific tolerance limits. For example, if a difference between these two trials is not greater than 12% difference in the raw scale, it is not considered a clinical influence or a scientific difference. Therefore, the clinically acceptable compliance interval must be Δ (-0.1133, +0.1133), which corresponds to a mismatch rate of 0.034. Conformity assessment based on this clinical interval is also presented in Figure 3, which shows that all differences are within clinically acceptable interval limits. With this clinically acceptable compliance interval, the discordanity rate is α = 0.034 and no difference is outside the clinically defined compliance interval, which quantifies the concordance between the new test and the current power measurement test with a probability of tolerance β = 0.66 for the α rate of inadequacy = 0.034. Note that the probability of tolerance β use of the clinically accepted compliance interval is lower than that of the agreement interval determined at a α = 0.05 mismatch rate, since a lower dosing rate is used α = 0.034.

This also reflects the similar relationship between significant level and power in a hypothesis testing environment. Haber M, Barnhart HX. Correspondence coefficients for permanent observers. Stat Methods Med Res 2006;15:255-71. Stine WW. Inter-observer relationship agreement. Psychol Bull 1989; 106:341-7. Different metrics can be defined to measure concordance in theory.

In practice, however, a simple and intuitive measure of the correspondence for each pair (X, Y) (i.e. within the individual between methods), where the measure X (Y) is the value to be reported and it is a transformation of the initial selection, as for example. B logscale observation. An obvious simple and intuitive starting point and a generally accepted metric is the difference between the measurements for each pair, and then a declaration of conformity by comparing the difference at a given interval Δ. . . .

This entry was posted in Geen categorie. Bookmark the permalink.