Table 5.
Factor | Bias | Error | ||||||
---|---|---|---|---|---|---|---|---|
Model 2A: NC=WIV | Model 2B: NC=FT | Model 3A: NC=WIV | Model 3B: NC=FT | |||||
Overall F-test | Summary | Overall F-test p-value | Summary | KW test p-value |
Summary | KW test p-value |
Summary | |
ARV | <0.001 | Highest estimated bias for DRV and RGV; many statistically significant pairwise differences. | 0.253 | -- | <0.001 | Many; highest (ranks of) errors for DRV and RGV. | <0.001 | {DRV,NVP} > {RGV,RTV,ZDV}, and DRV>MVC. |
Lab | <0.001 | Labs 6 and 7 have higher recoveries than most other labs. Lab 2 has lowest recovery(below 100%). | <0.001 | Labs 6 and 7 have higher recoveries than most other labs. Lab 2 has lowest recovery (below 100%). | <0.001 | {3, 6, 7} >all others. | <0.001 | {3, 6, 7} >all others, and 3>6. |
WIV category, L/M/H within analyte | <0.001 | Low & Medium > High | 0.099 | -- | 0.872 | -- | 0.563 | -- |
WIV category, “power of ten” | 0.005 | Despite p<0.05, no pairwise differences were statistically significant. | 0.907 | -- | 0.125 | -- | 0.941 | -- |
Round | 0.093 | -- | 0.413 | -- | <0.001 | Round 23 > 24 & 25. | 0.094 | -- |
KW = Kruskal-Wallis test. ARV = antiretroviral drug, WIV = weighed-in value. L/M/H = low/medium/high. NC = nominal concentration. FT = final target.