simAugData: A list consisting of simulated data, design, baseline...

In hbbr: Hierarchical Bayesian Benefit-Risk Assessment Using Discrete Choice Experiment

Description

Simulated response data and associated information from 100 respondents each choosing preference from 12 choice cards. Choice cards were randomly generated from 36 total choices.

Usage

data(simAugData)

Format

A list consisting of simulated response data from 100 subjects, design information, baseline profiles, and true part-worth matrix of the Augmented HBBR model framework.

brdtaAug

A data frame of 100x12 rows and 10 columns consists of simulated responses from 100 subjects each providing tradeoff responses to 12 choice pairs. The 1st column consists of subject id. The 2nd column contains binary responses (1 indicating 1st of the choice pair was selected, 0 indicating 2nd was selected). Remaining 8 columns contain the design matrix X taking values 0, 1, or -1; a value of 1 or -1 is used to indicate presence of an attribute level in the 1st choice or in the 2nd choice of the choice pair, respectively; a value of 0 is used to indicate absence of an attribute in the choice pair. See Details below for more about the discrete choice experiment that is coded as design matrix X.

design

A list of structure (b, r, bl, rl), where b and r indicate number of benefit and risk attributes, bl is a vector of integers of size b consisting number of levels within each benefit attribute; similarly rl is a vector of integers of size r consisting number of levels within each risk attribute.

Z

A data frame of size 100x3 consists of baseline characteristics from 100 subjects. The 1st column of Z is vector of 1, the 2nd column consists of standardized age, and the 3rd column indicates disease status at baseline (1= present, -1 = not present). The 1st column being a constant vector of 1 indicates that the part-worth matrix to be estimated, say Del-hat, would have the overall mean part-worth in the 1st column, while 2nd and 3rd columns would consists of estimates of additive components of part-worth due to age and disease status.

Del

A matrix of true part-worth values of dimension 8x3. This was used along with X and Z to generate the responses from 100 subjects captured in the 2nd column of brdtaSim.

Details

The simulated discrete choice experiment (DCE) included 3 benefit attributes (b=2): B1, B2 (say) and 2 risk attributes (r=2): R1, R2 (say). There were 3 levels for each of the benefit attributes ("Low", "Moderate", "High") (i.e. bl= rep(3,2)) and 3 levels for each of the risk attributes ("None", "Mild", "Severe") (i.e. rl = rep(3,2)). The DCE produced 36 distinct non-dominant choice pairs each with one benefit and one risk attribute. Panels (questionnaires) were generated with 12 (randomly) selected choice pairs per panel from the set of 36 choice pairs. Since the part-worth of various levels within each attribute are to be measured relatively to the part-worth of the 1st level of the attribute, columns for the 1st level of the attributes are not required. Thus, we have sum(bl)-b + sum(br)-r = 8 columns are needed to obtain information on the X matrix which are stored as the last 8 columns of brdtaAugSim.

Examples

data(simAugData)

hbbr documentation built on Oct. 30, 2019, 9:47 a.m.