Performance Enhancement

The package includes parallel processing for bootstrap methods. Unidirectional and bidirectional bootstrap sampling methods are the methods that benefit the most from parallel processing.

Usage

# Use parallel processing for faster bootstrap
result = RDSmean(
    x='income',
    data=rds_data,
    var_est='tree_uni1',
    resample_n=2000,
    n_cores=8  # Use 8 cores for parallel processing
)

Parallel processing is available for all bootstrap-based statistical functions:

RDSmean() with bootstrap variance estimation
RDStable() with bootstrap variance estimation
RDSlm() with bootstrap variance estimation

Performance Comparison

Performance Scaling
Cores	Bootstrap Samples	Standard Time	Parallel Time	Speedup
1	1000	120s	120s	1.0x
4	1000	120s	18s	6.7x
8	1000	120s	12s	10.0x

Examples

All estimation functions support the n_cores parameter:

# Parallel mean calculation
mean_result = RDSmean(
    x='age',
    data=rds_data,
    var_est='tree_uni1',
    resample_n=1000,
    n_cores=4
)

# Parallel table calculation
table_result = RDStable(
    x="Sex",
    y="Race",
    data=rds_data,
    var_est='tree_uni1',
    resample_n=1000,
    n_cores=4
)

# Parallel regression
regression_result = RDSlm(
    data=rds_data,
    formula="Age ~ Sex",
    var_est='tree_uni1',
    resample_n=1000,
    n_cores=4
)