datbrghtterms in that sessions are
replaced with (coarsened to) modules, each of which contains 4 sessions. Under the open-enrollment
protocol, clients could start at the beginning of any module. Each module represents one of four
topic areas that collect sessions focused on those topics. The data are configured to support model runs
using engine functions (dpgrowmm, dpgrowmult, ddpgrow).
datbrghtmodtermsN = 815 total measures for P = 299 subjects. Each entry contains a composite Beck Depression Inventory-II (BDI-II) score.
The BDI-II score is a sum acoss 21 four-level items (each scored 0 - 3) with a higher score signifying a greater level of depressive symptoms.
(1,2,...,299. Note: Participating clients are de-identified in this dataset.
N (e.g. (0,0,0,...,1,1,1,...) , either {0,1} with usual care (UC) = 0 and group cognitive behavioral therapy (CBT) = 1.
There are 140 clients with CBT = 1 (even though 132 of these actually attend sessions) in order to facilitate an intent-to-treat comparison.
N. There are 3 distinct time points or measurement waves. e.g. (0,3,6,0,3,6,0,0,3,,,,).
The first measure is at baseline when clients enrolled to BRIGHT and at two post-treatment follow-ups at 3 and 6 months with response rates of 86
n[g] subjects linked to each of g = 1, ..., (G = 4) CBT therapy groups. Each group is specialized
to its own multiple membership (MM) term for employment of engine function dpgrowmmult. The number of CBT clients for each group are
(n[1] = 17, n[2] = 18, n[3] = 19, n[4] = 78) for a total of Paff = 132 clients that attended sessions, as compared to the 140 assigned to the CBT arm.
subj.aff across groups for modeling all groups, together, in a single MM term with engine function dpgrowmm.
G, n[g] x S[g] multiple membership weight matrices that together map the P_aff = 132 affected subjects (in each element of the
subj.aff list) to their particular sessions attended within their assigned group. There are a total of 61 CBT modules allocated to the G CBT therapy groups
as S[1] = 9, S[2] = 10, S[3] = 10, S[4] = 32. The study was designed for clients to attend 16 sessions organized into 4 modules of 4 sessions each. The 4 modules were offered
on an open enrollment or rotating basis. Additional make-up sessions added within 2 modules resulted in the possibility for some clients to attend up to 20 sessions.
n = 132 x S = 61 matrix object that concatenates the matrix entries of the list object from W.subj.aff into a block-diagonal matrix
(with each group of clients and modules disjoint and non-communicating with the others) for modeling in a single MM term under function dpgrowmm.
n.tot = 299 x (S+1) = 62 matrix object that maps all clients to modules for use under engine function ddpgrow. The first column is an intercept filled with 1's.
The rows for column 2:62 sum to 1 for the first 132 clients (who attend CBT modules) in a multiple membership fashion. The rows for non-CBT clients are are constructed
as (1,0,0,..,0) such that no attendance is the hold-out category for identification.
G=4 CBT open enrollment groups with each entry capturing the number of modules for that group. This vector is used to allocate
modules to the set of base distributions chosen with option under the ddpgrow engine function; for example, option = c("car","mvn","car","ind") assigns
the noted distributional choices to the corresponding blocks of modules collected in numdose.
G=4 CBT open enrollment groups that provides a label for each module block under a distinct base distribution. These labels will be
used in the rendered plots using accessor functions associated to ddpgrow.
"mmcar" or "mmigrp" prior formulations for session effects. Each item contains a vector
that specifies sub-group membership for each block of sessions within the G=4 MM terms. A sub-group would collect modules that communicate with each other,
but not with the modules of other sub-groups. For these data, the modules within each CBT therapy group all communicate. Then each vector in group contains
the single value 1 of length equal to the number of sessions in the applicable therapy group. This object is input for engine function dpgrowmult in the
case it is desired to place all of the G = 4 MM terms under prior "mmcar". One may employ the appropriate subset of list entries for those
terms under which it is desired to employ prior "mmcar".
G = 4 groups, for the S = 61 CBT modules for modeling under engine function dpgrowmm.
S[g] x S[g] CAR adjacency matrix used to model prior association among effects for each MM term under prior "mmcar",
where S[g] are the number of effects for CBT therapy group g. One may employ the approach subset of adjacency matrices for those terms under which one
desires to specify prior "mmcar".
(S = 61) x (S = 61) matrix object encoding the dependence structure among modules that concatenates the entries of Omega into a block diagonal structure for use in dpgrowmm.
S. M. Paddock and T. D. Savitsky (2012) Bayesian Hierarchical Semiparametric Modeling of Longitudinal Post-treatment Outcomes from Open-enrollment Therapy Groups, soon to appear in: JRSS Series A (Statistics in Society).