Hello,
I am trying to fit a nlme (non linear mixed effect). I am using the SelfStart
function SSlogis. However the data in my hand contains few observations per subject
(4 or less), so the nlsList doesn't work... In this case I should fixe initial
parameter estimates. I remark that values of initial estimates have a greater effect
on the model fit (i.e. loglikelihood, AIC and also on the significance of some
parameter). Below the fit of the same model for different values of initial
parameters:
1- Initial parameters = c(3, 20, -3.732865)
>summary(fm.nlme)
Nonlinear mixed-effects model fit by maximum likelihood
Model: MTL ~ SSlogis(age, Asym, xmid, scal)
Data: data
AIC BIC logLik
7.996858 33.22420 2.001571
Random effects:
Formula: list(Asym ~ 1, xmid ~ 1)
Level: optid
Structure: Diagonal
Asym xmid Residual
StdDev: 0.2577518 0.004429942 0.1616483
Fixed effects: list(Asym ~ 1, xmid ~ 1, scal ~ 1)
Value Std.Error DF t-value p-value
Asym 1.414976 0.029927 282 47.28124 <.0001
xmid 23.828353 9.517606 282 2.50361 0.0129
scal -5.108539 4.047323 282 -1.26220 0.2079
Correlation:
Asym xmid
xmid 0.489
scal -0.642 -0.953
Standardized Within-Group Residuals:
Min Q1 Med Q3 Max
-3.09765727 -0.45088811 0.00807415 0.53313907 2.45621965
Number of Observations: 495
Number of Groups: 211
2- Initial parameters: c(1.339120 8.559620 -0.296747)
> summary(fm.nlme)
Nonlinear mixed-effects model fit by maximum likelihood
Model: MTL ~ SSlogis(age, Asym, xmid, scal)
Data: dataMedian
AIC BIC logLik
39.82897 65.05632 -13.91449
Random effects:
Formula: list(Asym ~ 1, xmid ~ 1)
Level: optid
Structure: Diagonal
Asym xmid Residual
StdDev: 0.239688 3.209776 0.162114
Fixed effects: list(Asym ~ 1, xmid ~ 1, scal ~ 1)
Value Std.Error DF t-value p-value
Asym 1.423446 0.0193297 282 73.64024 <.0001
xmid 11.894065 0.7795040 282 15.25850 <.0001
scal -1.082031 0.3130138 282 -3.45682 6e-04
Correlation:
Asym xmid
xmid 0.001
scal -0.196 -0.513
Standardized Within-Group Residuals:
Min Q1 Med Q3 Max
-3.122723638 -0.451811989 0.001009602 0.540504726 2.398619776
Number of Observations: 495
Number of Groups: 211
I think its a problem of local optimum......
Help!!!!
Many thanks,
Abderrahim
Dr Abderrahim Oulhaj
Oxford University
Department of Pharmacology
Mansfield Road
Oxford OX1 3QT
Tel: +44 1865 224098
Fax: +44 1865 224099
Email: [EMAIL PROTECTED]
[[alternative HTML version deleted]]
______________________________________________
[EMAIL PROTECTED] mailing list
https://www.stat.math.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
