Perform Leave-One-Out Sensitivity Analysis

Recalculates meta-analysis results leaving out each study one at a time to assess the influence of individual studies.

Usage

leave_one_out(meta_result = NULL, yi = NULL, sei = NULL, method = NULL)

Arguments

meta_result

A MetaResult object from meta_analysis()

yi

Numeric vector of effect estimates (optional if in meta_result)

sei

Numeric vector of standard errors (optional if in meta_result)

method

Character. Tau-squared estimation method for the leave-one-out refits: "DL", "REML", or "PM". If NULL (default), uses meta_result@metadata$method when available; otherwise defaults to "DL".

Value

A list with components:

results

Data frame with estimates when each study is excluded

influential_studies

Character vector of influential studies

n_influential

Number of influential studies

effect_measure

Effect measure used

model

Model type (fixed/random)

method

Tau-squared estimation method used

Examples

# Leave-one-out sensitivity analysis
yi <- log(c(0.75, 0.82, 0.68, 0.91, 0.77))
sei <- c(0.12, 0.15, 0.18, 0.14, 0.11)
meta_res <- meta_analysis(yi = yi, sei = sei, effect_measure = "hr")
loo <- leave_one_out(meta_res)
loo$results
#>   excluded_study   estimate       se  ci_lower ci_upper I2 estimate_display
#> 1        Study 1 -0.2349472 249.4553 -489.1584 488.6885  0        0.7906126
#> 2        Study 2 -0.2572550 231.2617 -453.5218 453.0073  0        0.7731710
#> 3        Study 3 -0.2104521 221.7938 -434.9184 434.4975  0        0.8102179
#> 4        Study 4 -0.2832901 197.4333 -387.2455 386.6790  0        0.7533013
#> 5        Study 5 -0.2415265 227.5301 -446.1924 445.7094  0        0.7854280
#>   ci_lower_display ci_upper_display pct_change
#> 1    3.640832e-213    1.716828e+212   4.296221
#> 2    1.091392e-197    5.477348e+196  -4.790668
#> 3    1.310196e-189    5.010342e+188  14.274094
#> 4    6.628305e-169    8.561205e+167 -15.395834
#> 5    1.663786e-194    3.707792e+193   1.616187
loo$influential_studies
#> [1] "Study 3" "Study 4"