Skip to contents

Generate sample data with unobserved variables (for CAM-UV)

Source: R/generate_camuv_sample.r
generate_camuv_sample.Rd

Generates a 6-variable nonlinear additive model with two unobserved variables, following the structure of the CAM-UV tutorial of the Python lingam package: u1 is an unobserved common cause of x3 and x4 (an unobserved backdoor path, UBP), and u2 is an unobserved intermediate variable on the path from x2 to x5 (an unobserved causal path, UCP). Only x0-x5 are returned as observed data.

Usage

generate_camuv_sample(n = 500L, seed = NULL)

Arguments

n

number of samples (default: 500)

seed

random seed (default: NULL, i.e. do not reset the RNG state)

Value

list with three elements:

  • data: data.frame of the 6 observed variables (x0-x5).

  • adjacency_matrix: the true 6x6 adjacency matrix among the observed variables, following the m[to, from] convention. Entries are 0/1 edge indicators (the causal functions are nonlinear), directly comparable to the output of lingam_camuv(). The x3-x4 (UBP) and x2-x5 (UCP) entries are NA, matching the convention used by lingam_camuv().

  • confounded_pairs: 2-column integer matrix of the variable pairs (1-based column positions) connected through an unobserved variable, usable as a test oracle for lingam_camuv()'s confounded_pairs.

Details

The data-generating process (all error terms e() are runif(n, -2.5, 2.5); the tutorial's per-call random constants are replaced by the fixed constants below for reproducibility; each column is standardized before being used as a cause, as in the tutorial):


u1 (latent) ~ e();  x0..x5 ~ e()
x3 = x3 + (u1 + 1.5)^2
x4 = x4 + (u1 + 1.2)^2
x1 = x1 + (x0 + 1.2)^2
x3 = x3 + (x0 - 1.5)^2
x4 = x4 + (x2 + 1.0)^2
u2 (latent) = (x2 - 1.2)^2 + e()
x5 = x5 + (u2 + 1.5)^2

Examples

confounded <- generate_camuv_sample(n = 200, seed = 1)
head(confounded$data)
#>           x0         x1         x2       x3        x4         x5
#> 1 -0.8716163 -0.3465264  0.5446793 1.942204 1.2914382 0.30235032
#> 2 -0.4753220 -0.2896798 -1.0796731 2.666484 0.4709917 2.26086593
#> 3  0.2707997  0.7383932  1.5577423 1.990869 4.3486061 0.85648319
#> 4  1.5173293  2.0465938  1.3639420 1.307915 2.7211432 0.73327140
#> 5 -1.1088636 -0.5210769  1.5211244 3.259314 2.3822889 0.04943462
#> 6  1.4808349  2.3919707  0.7668779 2.224420 2.6211662 0.35048326
confounded$adjacency_matrix
#>    x0 x1 x2 x3 x4 x5
#> x0  0  0  0  0  0  0
#> x1  1  0  0  0  0  0
#> x2  0  0  0  0  0 NA
#> x3  1  0  0  0 NA  0
#> x4  0  0  1 NA  0  0
#> x5  0  0 NA  0  0  0
confounded$confounded_pairs
#>      var1 var2
#> [1,]    3    6
#> [2,]    4    5