Diagnostics: R and Phoenix [Study As­sess­ment]

Hi ElMaestro et al.,

» Extract some model diagnostics: DF's and LogLikelihood, and compare to find out which result is the better candidate.

I can only provide the results of R and Phoenix:

R 3.2.5
library(nlme)
Subj   <- c(1, 2, 4, 5, 6, 4, 5, 6, 7, 8, 9, 7, 8, 9)
Dose   <- c(25, 25, 50, 50, 50, 250, 250, 250, 75, 75, 75, 250, 250, 250)
AUC    <- c(326.40, 437.82, 557.47, 764.85, 943.59, 2040.84, 2989.29,
4107.58, 1562.42, 982.02, 1359.68, 3848.86, 4333.10, 3685.55)
Cmax   <- c(64.82, 67.35, 104.15, 143.12, 243.63, 451.44, 393.45,
796.57, 145.13, 166.77, 296.90, 313.00, 387.00, 843.00)
resp   <- data.frame(Subj, Dose, Cmax, AUC)
resp$Subj <- factor(resp$Subj)
muddle <- lme(log(Cmax) ~ log(Dose), data=resp, random=~1|Subj)
sum.muddle <- summary(muddle)
CI.muddle  <- intervals(muddle, level=0.9, which="fixed")
print(sum.muddle); CI.muddle\$fixed[, ]
Linear mixed-effects model fit by REML
Data: resp
AIC      BIC    logLik
14.24355 16.18317 -3.121774

Random effects:
Formula: ~1 | Subj
(Intercept)  Residual
StdDev:   0.3347319 0.1206792

Fixed effects: log(Cmax) ~ log(Dose)
Value  Std.Error DF   t-value p-value
(Intercept) 1.9413858 0.24314072  7  7.984618   1e-04
log(Dose)   0.7617406 0.04727976  5 16.111347   0e+00
Correlation:
(Intr)
log(Dose) -0.863

Standardized Within-Group Residuals:
Min          Q1         Med          Q3         Max
-1.07547728 -0.35579449 -0.03301391  0.45088601  0.91853654

Number of Observations: 14
Number of Groups: 8
lower      est.     upper
(Intercept) 1.4807366 1.9413858 2.4020350
log(Dose)   0.6664696 0.7617406 0.8570116

Phoenix 6.47.0.768
Model Specification and User Settings
Dependent variable : logCmax
Transform : None
Fixed terms : int+logDose
Random/repeated terms : Subject
Denominator df option : satterthwaite

Class variables and their levels
Subject :    1   2   4   5   6   7   8   9

Final variance parameter estimates:
Var(Subject)    0.112045
Var(Residual)    0.0145635

REML log(likelihood)   -0.623363
-2* REML log(likelihood)    1.24673
Akaike Information Crit.    9.24673
Schwarz Bayesian Crit.   11.1864

Effect:Level Estimate   StdError Denom_DF  T_stat  P_value Conf T_crit  Lower_CI  Upper_CI
---------------------------------------------------------------------------------------------
int  1.9413858 0.2431407   9.2    7.98462 1.980E-5   90  1.829 1.4967592 2.3860125
logDose:logDose 0.7617406 0.0472798   5.9   16.11135 4.241E-6   90  1.949 0.6695783 0.8539029

Estimates and their SEs are exactly the same. CIs are not (due to different DFs?).

PS: An ideas how to weight by 1/log(Dose) in lme()? Suggested by Chow/Liu and gives me a better fit in Phoenix.

Dif-tor heh smusma 🖖
Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes Ing. Helmut Schütz 