Benchmark Mixed Model Implementations

Compare performance between the standard mixed_solve implementation and the optimized mixed_solve_optimized version.

Usage

benchmark_mixed_solve(X, Z, K = NULL, Y, n_reps = 5)

Arguments

X

Fixed effects design matrix

Z

Random effects design matrix

K

Kinship matrix (optional, defaults to identity)

Y

Response matrix (n x V)

n_reps

Number of repetitions for benchmarking

Value

Data frame with timing results

Examples

# \donttest{
X <- matrix(rnorm(100 * 2), 100, 2)
Z <- matrix(rnorm(100 * 3), 100, 3)
Y <- matrix(rnorm(100 * 5), 100, 5)
benchmark_mixed_solve(X, Z, Y = Y, n_reps = 2)
#> Benchmarking mixed model implementations...
#> Data: n = 100 , p = 2 , q = 3 , voxels = 5 
#> Testing standard mixed_solve...
#> Testing optimized mixed_solve...
#> 
#> Results:
#>      method   mean_time median_time    min_time    max_time      sd_time
#> 1  standard 0.005525112 0.005525112 0.005077124 0.005973101 0.0006335514
#> 2 optimized 0.017022133 0.017022133 0.010108948 0.023935318 0.0097767202
#> 
#> Per-voxel timing:
#> Standard: 0.0011 sec/voxel
#> Optimized: 0.0034 sec/voxel
#> Speedup: 0.32x 
#>      method   mean_time median_time    min_time    max_time      sd_time
#> 1  standard 0.005525112 0.005525112 0.005077124 0.005973101 0.0006335514
#> 2 optimized 0.017022133 0.017022133 0.010108948 0.023935318 0.0097767202
# }