A small balanced Gaussian hierarchical model dataset for quick testing and demonstration. Contains J = 10 groups with n_j = 20 observations each (N = 200 total), equal weights (DEFF = 1), and two fixed-effect covariates (intercept + within-cluster covariate).
Format
A list with components:
- draws
Matrix of posterior draws (4000 x 12), columns 1:2 are fixed effects (beta), columns 3:12 are random effects (theta).
- y
Continuous outcome vector (length 200).
- X
Design matrix (200 x 2) with columns: intercept, x_within.
- group
Integer group indicator (1 to 10).
- weights
Survey weights (all 1.0, length 200).
- psu
PSU indicators (same as group).
- param_types
Character vector of length 2:
c("fe_between", "fe_within").- family
Model family:
"gaussian".- sigma_theta
Random effect SD (0.5).
- sigma_e
Residual SD (1.0).
- N
Number of observations (200).
- J
Number of groups (10).
- p
Number of fixed effects (2).
- B_ref
Analytical shrinkage factor (5/6).
- deff_ref
Reference design effect (1.0).
Details
With equal weights the design effect is 1.0, so DER values should be close to 1.0 across all parameters. This dataset is useful for verifying that the pipeline correctly identifies the absence of design effects.
Examples
data(sim_hlr)
str(sim_hlr, max.level = 1)
#> List of 15
#> $ draws : num [1:4000, 1:12] 2.09 2.05 2.11 1.98 1.93 ...
#> ..- attr(*, "dimnames")=List of 2
#> $ y : num [1:200] 1.173 1.73 2.698 1.094 0.951 ...
#> $ X : num [1:200, 1:2] 1 1 1 1 1 1 1 1 1 1 ...
#> ..- attr(*, "dimnames")=List of 2
#> $ group : int [1:200] 1 1 1 1 1 1 1 1 1 1 ...
#> $ weights : num [1:200] 1 1 1 1 1 1 1 1 1 1 ...
#> $ psu : int [1:200] 1 1 1 1 1 1 1 1 1 1 ...
#> $ param_types: chr [1:2] "fe_between" "fe_within"
#> $ family : chr "gaussian"
#> $ sigma_theta: num 0.5
#> $ sigma_e : num 1
#> $ N : int 200
#> $ J : int 10
#> $ p : int 2
#> $ B_ref : num 0.833
#> $ deff_ref : num 1