search

resplab / epicR

R package for the Evaluation Platform in COPD (EPIC), an agent-based whole-disease model for projection of health and economic outcomes and COPD interventions.
11 stars 11 forks source link

Modified Zafar's CMAJ Regression #32

Closed aminadibi closed 7 years ago

aminadibi commented 7 years ago

Modified Zafar's CMAJ Regression (removing O'Connor's slope) yields these new coefficients:

Baseline FEV1 (mL)
Effect Estimate Standard Error DF t Value Pr > t
Intercept 1427.5 1400.5 5583 1.02 0.3081
age_baseline -5.08 6.231 2.40E+04 -0.82 0.4146
gender 482.5 13.31 2.40E+04 36.26 <.0001
weight (kg) -0.49 0.391 2.40E+04 -1.25 0.2116
height (m) -1872.5 1540.7 2.40E+04 -1.22 0.2242
ht_squared (m2) 1951.3 444.1 2.40E+04 4.39 <.0001
smoker -92.21 12.62 2.40E+04 -7.31 <.0001
age_basel*ht_squared -8.32 2.077 2.40E+04 -4 <.0001
Rate of FEV1 Declime (mL/yr)
Effect Estimate Standard Error DF t Value Pr > t
Intercept -154.3 143.3 5585 -1.08 0.2815
age_baseline 2.344 0.644 2.40E+04 3.64 0.0003
gender -7.62 1.374 2.40E+04 -5.55 <.0001
weight (kg) 0.126 0.041 2.40E+04 3.11 0.0019
height (m) 58.35 157.9 2.40E+04 0.37 0.7117
ht_squared (m2) 18.07 45.58 2.40E+04 0.4 0.6918
Smoker -30.74 1.242 2.40E+04 -24.76 <.0001
baseline_age * height sq -0.93 0.215 2.40E+04 -4.34 <.0001
yearYear -1.46 0.064 2.40E+04 -22.76 <.0001

Model fit statistics are:

Number of Observations
Number of Observations Read 35046
Estimated G Matrix
Row Effect id Col1 Col2
1 Intercept A000091 0.1006 0.00087
2 year A000091 0.00087 0.000759
Fit Statistics
-2 Res Log Likelihood -15198.5
AIC (smaller is better) -15190.5
AICC (smaller is better) -15190.5
BIC (smaller is better) -15164
Null Model Likelihood Ratio Test
DF Chi-Square Pr > ChiSq
3 45661.26 <.0001
aminadibi commented 7 years ago

One could also model Percent Predicted FEV1, which would not require height and weight coefficients:

Baseline Percent Predicted FEV1
Effect Estimate Standard Error DF t Value Pr > t
Intercept 865.8 9.108 5588 95.06 <.0001
age_baseline -2.5 0.168 2.40E+04 -14.91 <.0001
gender 34.17 2.362 2.40E+04 14.46 <.0001
smoker -57.88 3.277 2.40E+04 -17.66 <.0001
Percent Predicted FEV1 Decline
Effect Estimate Standard Error DF t Value Pr > t
Intercept 21.13 0.432 5589 48.92 <.0001
age_baseline -0.07 7.69E-03 2.40E+04 -9.35 <.0001
gender 0.802 0.107 2.40E+04 7.47 <.0001
Smoker -3.8 0.142 2.40E+04 -26.8 <.0001
yearYear -1.83 0.014 2.40E+04 -130.91 <.0001

with model fit statistics as follows:

Fit Statistics
-2 Res Log Likelihood -123232
AIC (smaller is better) -123226
AICC (smaller is better) -123226
BIC (smaller is better) -123206
Null Model Likelihood Ratio Test
DF Chi-Square Pr > ChiSq
2 52973.77 <.0001