Skip to main content
. 2016 Nov;51(11):849–857. doi: 10.4085/1062-6050-51.9.09

Table 4.

Coefficients from the Best-Fitting Logistic Regression Model Predicting Overuse Musculoskeletal Injurya

Step
Coefficients
Estimate
Standard Error
Z Value
P Value
Δ Deviance
Akaike Information Criterion
1 Intercept −1.983 0.096 −20.76 <.001 0 1343.5
2 Point of instruction 1 0.026 0.309 0.08 .93 11.2 1338.3
2 Point of instruction 2 0.683 0.330 2.07 .04
2 Point of instruction 3 0.152 0.329 0.46 .64
3b ln (Run.c) 1.972 0.423 4.66 <.001 26.4 1313.9
4 FitSum.c −0.006 0.004 −1.44 .15 1.9 1314.0
5 Age.c 0.028 0.022 1.29 .20 1.3 1314.4

Abbreviations: Age.c, centered mean of ages; Δ Deviance, change in deviance compared with the model in the previous step; FitSum.c, centered mean of the push-up and sit-up variables combined; ln (Run.c), centered mean of the natural logarithm of run time.

a 

Points of instruction 1 through 3 are simple contrast codes for the different programs of instruction (the intercept is thus equivalent to the grand mean). Null deviance = 1643.0 with 1787 degrees of freedom. Adding the interactions of ln (Run.c), FitSum.c, and Age.c in step 6 (not shown) did not improve the fit of the model.

b 

Denotes the best-fitting model according to Akaike information criterion and Δ Deviance (P < .05).