This function computes a threshold for indirect or direct effect estimates that controls the false discovery rate according to estimates made using real and synthetic null data, against the null hypotheses that effects are zero. It computes the proportion of synthetic null estimates that are among the top K largest effects (in magnitude) as an estimate of the FDR.
Arguments
- contrast
A data.frame summarizing the differences between outcomes across hypothetical treatments, typically as output by
null_contrast. Each row is one outcome in one hypothetical scenario.- effect
Either "indirect_overall" (the default), "indirect_pathwise", or "direct_effect" specifying the type of effect that we want to select.
- q_value
The target for false discovery rate control. The last time the estimated FDR is above this threshold is smallest magnitude of effect size that we will consider.
Examples
# example with null data - notice synthetic data has larger effect.
exper <- demo_joy() |>
mediation_data("PHQ", "treatment", starts_with("ASV"))
multimedia(exper) |>
estimate(exper) |>
null_contrast(exper) |>
fdr_summary("direct_effect")
#> Fitting the nullified model...
#> Generating synthetic data...
#> Fitting the full model on synthetic data...
#> Estimating effects on real and synthetic data...
#> Warning: There was 1 warning in `summarise()`.
#> ℹ In argument: `ix = max(.data$rank)`.
#> Caused by warning in `max()`:
#> ! no non-missing arguments to max; returning -Inf
#> # A tibble: 2 × 6
#> source outcome direct_effect rank fdr_hat keep
#> <chr> <chr> <dbl> <int> <dbl> <lgl>
#> 1 synthetic PHQ 0.398 1 1 FALSE
#> 2 real PHQ -0.298 2 0.5 FALSE
multimedia(exper) |>
estimate(exper) |>
null_contrast(exper, "M->Y", indirect_overall) |>
fdr_summary("indirect_overall")
#> Fitting the nullified model...
#> Generating synthetic data...
#> Fitting the full model on synthetic data...
#> Estimating effects on real and synthetic data...
#> Warning: There was 1 warning in `summarise()`.
#> ℹ In argument: `ix = max(.data$rank)`.
#> Caused by warning in `max()`:
#> ! no non-missing arguments to max; returning -Inf
#> # A tibble: 2 × 6
#> source outcome indirect_effect rank fdr_hat keep
#> <chr> <chr> <dbl> <int> <dbl> <lgl>
#> 1 synthetic PHQ -0.120 1 1 FALSE
#> 2 real PHQ -0.0447 2 0.5 FALSE
# example with another dataset - synthetic effect is smaller.
exper <- demo_spline(tau = c(2, 1)) |>
mediation_data(starts_with("outcome"), "treatment", "mediator")
multimedia(exper) |>
estimate(exper) |>
null_contrast(exper) |>
fdr_summary("direct_effect")
#> Fitting the nullified model...
#> Generating synthetic data...
#> Fitting the full model on synthetic data...
#> Estimating effects on real and synthetic data...
#> # A tibble: 4 × 6
#> source outcome direct_effect rank fdr_hat keep
#> <chr> <chr> <dbl> <int> <dbl> <lgl>
#> 1 real outcome_1 -1.72 1 0 TRUE
#> 2 real outcome_2 -0.847 2 0 FALSE
#> 3 synthetic outcome_2 0.0259 3 0.333 FALSE
#> 4 synthetic outcome_1 -0.00569 4 0.5 FALSE
multimedia(exper) |>
estimate(exper) |>
null_contrast(exper, "M->Y", indirect_overall) |>
fdr_summary("indirect_overall")
#> Fitting the nullified model...
#> Generating synthetic data...
#> Fitting the full model on synthetic data...
#> Estimating effects on real and synthetic data...
#> # A tibble: 4 × 6
#> source outcome indirect_effect rank fdr_hat keep
#> <chr> <chr> <dbl> <int> <dbl> <lgl>
#> 1 real outcome_2 1.01 1 0 TRUE
#> 2 real outcome_1 0.772 2 0 FALSE
#> 3 synthetic outcome_2 -0.0213 3 0.333 FALSE
#> 4 synthetic outcome_1 -0.0183 4 0.5 FALSE