Table 3.
Reordered comparisons reveal correct p values for Tukey’s post-hoc tests
| Test | Ottenhoff et al. (incorrect order) | Muscas et al. (corrected order) | ||||
|---|---|---|---|---|---|---|
| Estimate | z value | p value | Estimate | z value | p value | |
| WT, Veh vs lova | 0.1542 | 0.168 | 1.0000 | 0.1542 | 0.168 | 1.0000 |
| WT, simvalow vs Veh | –0.4700 | –0.366 | 0.9997 | –0.2288 | –0.196 | 1.0000 |
| WT, simvahigh vs Veh | –0.3830 | –0.297 | 0.9999 | –0.3159 | –0.271 | 0.9999 |
| WT, simvalow vs lova | –0.3159 | –0.271 | 0.9999 | –0.3830 | –0.297 | 0.9999 |
| WT, simvahigh vs lova | –0.2288 | –0.196 | 1.0000 | –0.4700 | –0.366 | 0.9997 |
| WT, simvalow vs simvahigh | –0.0870 | –0.059 | 1.0000 | –0.0870 | –0.059 | 1.0000 |
| KO, Veh vs lova | –2.1016 | –2.872 | 0.0406 | –2.1016 | –2.872 | 0.0406 |
| KO, simvalow vs Veh | 1.4816 | 1.666 | 0.5570 | 0.2963 | 0.397 | 0.9995 |
| KO, simvahigh vs Veh | 2.3979 | 2.573 | 0.0932 | –0.6200 | –0.897 | 0.9607 |
| KO, simvalow vs lova | –0.6200 | –0.897 | 0.9607 | 2.3979 | 2.573 | 0.0932 |
| KO, simvahigh vs lova | 0.2963 | 0.397 | 0.9995 | 1.4816 | 1.666 | 0.5570 |
| KO, simvalow vs simvahigh | –0.9163 | –1.017 | 0.9288 | –0.9163 | –1.017 | 0.9288 |
The regression model R script used by Ottenhoff et al. (2020) assigns different functions to set up the regression model matrix (“unique”) versus the Tukey’s contrast matrix (“tables”). This results in different order of groups for the two matrices, which results in assignment of different headings to the test results. An altered version of the script with the factors level set in the same order for the model matrix and contrast matrix shows the correct Tukey’s test results (see Extended Data Figure 1-3). Estimate and z value are multiplied by –1 to reflect the corresponding tests headings. Reversed values are italicized and the corrected p values reported by Ottenhoff are in bold.