medRCT

https://github.com/t0ngchen/medrct
Admin
medRCT is an R package for causal mediation analysis that estimates interventional effects mapped to...
Tokens:5,704
Snippets:63
Trust Score:5.7
Update:6 months ago
Show doc for...
Tokens:
Raw
### Install medRCT from GitHub

Source: https://github.com/t0ngchen/medrct/blob/master/README.md

Use the remotes package to install the latest stable version of medRCT from the GitHub repository.

```r
remotes::install_github("T0ngChen/medRCT")
```

--------------------------------

### Initialize medRCT Analysis

Source: https://github.com/t0ngchen/medrct/blob/master/README.md

Load the package and set a random seed to ensure reproducibility of the simulation results.

```r
# Load the medRCT package
library(medRCT)

# Set a seed for reproducibility
set.seed(2025)
```

--------------------------------

### Configure Bootstrap Parameters

Source: https://context7.com/t0ngchen/medrct/llms.txt

Adjust bootstrap resampling settings or disable bootstrapping to obtain point estimates only.

```r
# Custom bootstrap configuration with percentile CIs
med_boot_perc <- medRCT(
  dat = LSACdata,
  exposure = "sep",
  outcome = "child_mh",
  mediators = c("parent_mh", "preschool_att"),
  intermediate_confs = "fam_stress",
  confounders = c("child_sex", "child_atsi", "mat_cob", "mat_engl", "mat_age"),
  intervention_type = "shift_k",
  effect_measure = "RD",
  bootstrap = TRUE,
  boot_args = list(
    R = 500,              # Number of bootstrap replicates
    stype = "i",          # Bootstrap statistic type
    ci.type = "perc"      # Percentile confidence intervals
  ),
  mcsim = 200
)

# Run without bootstrapping for point estimates only
med_no_boot <- medRCT(
  dat = LSACdata,
  exposure = "sep",
  outcome = "child_mh",
  mediators = c("parent_mh", "preschool_att"),
  intermediate_confs = "fam_stress",
  confounders = c("child_sex", "child_atsi", "mat_cob", "mat_engl", "mat_age"),
  intervention_type = "shift_k",
  effect_measure = "RD",
  bootstrap = FALSE,      # No bootstrapping - point estimates only
  mcsim = 200
)

# View point estimates only
med_no_boot$est
```

--------------------------------

### Load and Explore LSAC Dataset

Source: https://context7.com/t0ngchen/medrct/llms.txt

Load the included simulated dataset and inspect its structure.

```r
# Load and explore the included dataset
data(LSACdata)

# Dataset structure
str(LSACdata)
#> 'data.frame':	5107 obs. of  11 variables:
#>  $ child_sex     : num  0 NA NA NA 1 1 ...
#>  $ child_atsi    : num  1 0 0 0 0 0 ...
#>  $ mat_cob       : num  0 0 0 0 0 0 ...
#>  $ mat_engl      : num  0 0 0 0 0 0 ...
#>  $ mat_age       : num  1 NA NA NA 1 1 ...
#>  $ sep           : num  0 0 0 0 1 0 ...        # Exposure: socioeconomic position
#>  $ fam_stress    : num  0 NA NA NA 0 1 ...     # Intermediate confounder
#>  $ parent_mh     : num  0 NA NA NA 0 1 ...     # Mediator 1: parental mental health
#>  $ preschool_att : num  1 0 0 0 0 0 ...        # Mediator 2: preschool attendance
#>  $ child_mh      : num  0 0 1 0 1 1 ...        # Binary outcome: child mental health
#>  $ child_SDQscore: num  8.92 7.35 12.82 ...    # Continuous outcome: SDQ score
```

--------------------------------

### Launch Interactive Model Assessment

Source: https://context7.com/t0ngchen/medrct/llms.txt

Use the Shiny application to validate regression models and configure parameters before running the full analysis.

```r
# Launch Shiny app for model assessment
if (interactive()) {
  medRCT_shiny(data = LSACdata)
}
```

--------------------------------

### Inspect Dataset and Define Confounders

Source: https://github.com/t0ngchen/medrct/blob/master/README.md

View the initial rows of the dataset and define the vector of confounders to be used in the analysis.

```R
head(LSACdata)

# Define confounders for the analysis
confounders <- c("child_sex", "child_atsi", "mat_cob", "mat_engl", "mat_age")
```

--------------------------------

### Simple Mediation Analysis Without Intermediate Confounders

Source: https://context7.com/t0ngchen/medrct/llms.txt

Performs mediation analysis when there are no exposure-induced mediator-outcome confounders. Set 'intermediate_confs' to NULL. Requires specifying data, exposure, outcome, mediators, confounders, and intervention type.

```r
med_simple <- medRCT(
  dat = LSACdata,
  exposure = "sep",
  outcome = "child_mh",
  mediators = c("parent_mh", "preschool_att"),
  intermediate_confs = NULL,          # No intermediate confounders
  confounders = c("child_sex", "child_atsi", "mat_cob", "mat_engl", "mat_age"),
  intervention_type = "all",
  effect_measure = "RD",
  bootstrap = TRUE,
  boot_args = list(R = 100, stype = "i", ci.type = "norm"),
  mcsim = 200
)
```

```r
summary(med_simple)
```

--------------------------------

### Analyze Continuous Outcome with Shift-k Intervention

Source: https://context7.com/t0ngchen/medrct/llms.txt

Analyzes a continuous outcome using the difference in means as the effect measure, with a 'shift_k' intervention type. Requires specifying data, exposure, outcome, mediators, intermediate confounders, confounders, and effect measure.

```r
med_continuous <- medRCT(
  dat = LSACdata,
  exposure = "sep",
  outcome = "child_SDQscore",         # Continuous outcome variable
  mediators = c("parent_mh", "preschool_att"),
  intermediate_confs = "fam_stress",
  confounders = c("child_sex", "child_atsi", "mat_cob", "mat_engl", "mat_age"),
  intervention_type = "shift_k",
  effect_measure = "Diff",            # Difference in means for continuous outcome
  bootstrap = TRUE,
  boot_args = list(R = 100, stype = "i", ci.type = "norm"),
  mcsim = 200
)
```

```r
summary(med_continuous)
```

--------------------------------

### Customize Exposure-Confounder Interactions

Source: https://context7.com/t0ngchen/medrct/llms.txt

Control the inclusion of exposure-confounder interactions in regression models using the interactions_XC argument.

```r
# No exposure-confounder interactions
med_no_int <- medRCT(
  dat = LSACdata,
  exposure = "sep",
  outcome = "child_mh",
  mediators = c("parent_mh", "preschool_att"),
  intermediate_confs = "fam_stress",
  confounders = c("child_sex", "child_atsi", "mat_cob", "mat_engl", "mat_age"),
  interactions_XC = "none",           # Remove all exposure-confounder interactions
  intervention_type = "shift_k",
  effect_measure = "RD",
  bootstrap = TRUE,
  mcsim = 200
)

# Custom interaction specification
med_custom_int <- medRCT(
  dat = LSACdata,
  exposure = "sep",
  outcome = "child_mh",
  mediators = c("parent_mh", "preschool_att"),
  intermediate_confs = "fam_stress",
  confounders = c("child_sex", "child_atsi", "mat_cob", "mat_engl", "mat_age"),
  interactions_XC = "sep:child_sex + sep:mat_age",  # Specific interactions only
  intervention_type = "shift_k",
  effect_measure = "RD",
  bootstrap = TRUE,
  mcsim = 200
)
```

--------------------------------

### Perform Single Mediator Analysis

Source: https://context7.com/t0ngchen/medrct/llms.txt

Execute a mediation analysis with a single mediator using the shift_k intervention type.

```r
med_single <- medRCT(
  dat = LSACdata,
  exposure = "sep",
  outcome = "child_mh",
  mediators = "parent_mh",            # Single mediator
  intermediate_confs = "fam_stress",
  confounders = c("child_sex", "child_atsi", "mat_cob", "mat_engl", "mat_age"),
  intervention_type = "shift_k",      # Only option for single mediator
  effect_measure = "RD",
  bootstrap = TRUE,
  boot_args = list(R = 100, stype = "i", ci.type = "norm"),
  mcsim = 200
)
#> Only able to estimate the effect type 'shift_k' with a single mediator.

summary(med_single)
```

--------------------------------

### Load Data and Estimate Interventional Effects with medRCT

Source: https://context7.com/t0ngchen/medrct/llms.txt

Loads the LSACdata dataset and estimates various interventional mediation effects using the medRCT function. Requires specifying confounders, mediators, and intervention types. Bootstrapping is enabled for confidence intervals.

```r
library(medRCT)

# Load included simulated dataset based on Longitudinal Study of Australian Children
data(LSACdata)
head(LSACdata)
#>   child_sex child_atsi mat_cob mat_engl mat_age sep fam_stress parent_mh
#> 1         0          1       0        0       1   0          0         0
#> 2        NA          0       0        0      NA   0         NA        NA
#> 3        NA          0       0        0      NA   0         NA        NA
#> 4        NA          0       0        0      NA   0         NA        NA
#> 5         1          0       0        0       1   1          0         0
#> 6         1          0       0        0       1   0          1         1
#>   preschool_att child_mh child_SDQscore
#> 1             1        0       8.924660
#> 2             0        0       7.349826
#> 3             0        1      12.824643
#> 4             0        0       6.611369
#> 5             0        1      10.329341
#> 6             0        1      13.552515

# Define baseline confounders
confounders <- c("child_sex", "child_atsi", "mat_cob", "mat_engl", "mat_age")

# Define intermediate (exposure-induced) confounders
intermediate_confs <- "fam_stress"

# Set seed for reproducibility
set.seed(2025)

# Estimate all types of interventional effects
med_res <- medRCT(
  dat = LSACdata,
  exposure = "sep",                              # Binary exposure: socioeconomic position
  outcome = "child_mh",                          # Binary outcome: child mental health problems
  mediators = c("parent_mh", "preschool_att"),   # Two mediators of interest
  intermediate_confs = intermediate_confs,       # Exposure-induced confounder
  confounders = confounders,                     # Baseline confounders
  interactions_XC = "all",                       # Include all exposure-confounder interactions
  intervention_type = "all",                     # Estimate all three effect types
  effect_measure = "RD",                         # Risk difference for binary outcome
  bootstrap = TRUE,                              # Enable bootstrapping for CIs
  boot_args = list(R = 200, stype = "i", ci.type = "norm"),
  mcsim = 200                                    # Monte Carlo simulations
)
#> Assumed causal order for estimating effect of type 'shift_k_order': parent_mh, preschool_att
#> Conducting complete case analysis, 2499 observations were excluded due to missing data.

```

--------------------------------

### Estimate Interventional Effects

Source: https://github.com/t0ngchen/medrct/blob/master/README.md

Run the medRCT function to estimate interventional effects. Note that incomplete records are removed by default and a higher number of Monte Carlo simulations is recommended for production.

```R
med_res <- medRCT(
  dat = LSACdata,                      
  exposure = "sep",                    
  outcome = "child_mh",                
  mediators = c("parent_mh", "preschool_att"), 
  intermediate_confs = "fam_stress",  # intermediate confounders 
  confounders = confounders,           
  bootstrap = TRUE,                    
  intervention_type = "shift_k",       
  mcsim = 50                          
)
```

--------------------------------

### Estimate Shift-k Intervention Effects

Source: https://context7.com/t0ngchen/medrct/llms.txt

Estimates interventional effects for shifting individual mediator distributions without considering flow-on effects to other mediators. This is faster for computation. Requires specifying data, exposure, outcome, mediators, intermediate confounders, confounders, and intervention type.

```r
med_shift_k <- medRCT(
  dat = LSACdata,
  exposure = "sep",
  outcome = "child_mh",
  mediators = c("parent_mh", "preschool_att"),
  intermediate_confs = "fam_stress",
  confounders = c("child_sex", "child_atsi", "mat_cob", "mat_engl", "mat_age"),
  intervention_type = "shift_k",     # Only estimate shift_k effects
  effect_measure = "RD",
  bootstrap = TRUE,
  boot_args = list(R = 100, stype = "i", ci.type = "norm"),
  mcsim = 200
)
```

```r
summary(med_shift_k)
```

--------------------------------

### Calculate Risk Ratio Effect Measure

Source: https://context7.com/t0ngchen/medrct/llms.txt

Calculates the risk ratio as the effect measure for binary outcomes. This is an alternative to risk difference. Requires specifying data, exposure, outcome, mediators, intermediate confounders, confounders, and effect measure.

```r
med_rr <- medRCT(
  dat = LSACdata,
  exposure = "sep",
  outcome = "child_mh",
  mediators = c("parent_mh", "preschool_att"),
  intermediate_confs = "fam_stress",
  confounders = c("child_sex", "child_atsi", "mat_cob", "mat_engl", "mat_age"),
  intervention_type = "shift_k",
  effect_measure = "RR",              # Risk ratio for binary outcome
  bootstrap = TRUE,
  boot_args = list(R = 100, stype = "i", ci.type = "norm"),
  mcsim = 200
)
```

```r
summary(med_rr)
```

--------------------------------

### View Mediation Analysis Results

Source: https://context7.com/t0ngchen/medrct/llms.txt

Displays a summary of the mediation analysis results, including estimated indirect effects, direct effects, and expected outcomes in different trial arms. This is typically used after running a mediation analysis model.

```r
summary(med_res)
```
Tokens:
Raw