Hi, Gang.

I want to use 3dLME to analyze data with family dependencies, similar to Example 6 in Chen et al (2013). However, I can't seem to map the explanation in the paper onto the command syntax of 3dLME.

As a starting point, I just want to replicate a model I can run using lme. The experiment involves 20 subjects organized as 10 families (Fam), each comprised of 2 siblings (Sib), each of whom receives 2 experimental conditions (Cond [emot and shape]).

Here's the data structure (formulated as a data table for 3dLME):

Subj	Fam	Sib	Cond	InputFile	\
s01	0	0	shape	TEST_Fam0_Sib0_S.nii.gz	\
s01	0	0	emot	TEST_Fam0_Sib0_E.nii.gz	\
s02	0	1	shape	TEST_Fam0_Sib1_S.nii.gz	\
s02	0	1	emot	TEST_Fam0_Sib1_E.nii.gz	\
s03	1	0	shape	TEST_Fam1_Sib0_S.nii.gz	\
...
s18	8	1	emot	TEST_Fam8_Sib1_E.nii.gz	\
s19	9	0	shape	TEST_Fam9_Sib0_S.nii.gz	\
s19	9	0	emot	TEST_Fam9_Sib0_E.nii.gz	\
s20	9	1	shape	TEST_Fam9_Sib1_S.nii.gz	\
s20	9	1	emot	TEST_Fam9_Sib1_E.nii.gz	\

Using lme, I might run an analysis (at a single voxel) to look at the main effect of Condition like this (where Bold is the data vector):

> lfit_ml <- lme(Bold ~ as.factor(Cond), random = ~1 | Fam/Sib, data=df2, method="ML")
> summary(lfit_ml)
Linear mixed-effects model fit by maximum likelihood
 Data: df2 
       AIC      BIC    logLik
  100.1961 108.6405 -45.09807

Random effects:
 Formula: ~1 | Fam
        (Intercept)
StdDev:  0.04589337

 Formula: ~1 | Sib %in% Fam
         (Intercept)  Residual
StdDev: 1.866988e-05 0.7457489

Fixed effects: Bold ~ as.factor(Cond) 
                    Value Std.Error DF t-value p-value
(Intercept)      5.030772 0.1717333 19 29.2941  0.0000
as.factor(Cond)2 0.753054 0.2419529 19  3.1124  0.0057
 Correlation: 
                 (Intr)
as.factor(Cond)2 -0.704

Standardized Within-Group Residuals:
       Min         Q1        Med         Q3        Max 
-2.2115801 -0.6770179 -0.0374632  0.8460666  2.0861792 

Number of Observations: 40
Number of Groups: 
         Fam     Sib %in% Fam 
            10             20 
>

Can you propose the syntax for a 3dLME model that would replicate this kind of analysis?

Many thanks in advance,
Ruskin