Run PIRR-style modeling and summary plots for multiple binary outcomes
Source:R/poisson_analysis.R
pirr_results.RdThis function runs Poisson regression models using splines for age and compares effects over two time-related variables (`caika` and `cever`) for each binary outcome variable found in the dataset. It returns model-based predictions, summary tables, and plots for each outcome.
Arguments
- adat
A data frame containing the data. Must include columns `Age`, `caika`, `pyrs`, `cever`, and at least one binary outcome variable (e.g., disease flags). Created by function `pirr_data()`.
- colors
A character vector of two hex colors for plotting binary outcome levels in the age vs exposure plots. Default is
c("#5BC0DE", "#D9534F").- limits
A numeric vector of two values specifying the y-axis limits (log10 scale) for the SIR plots. Default is
c(0.3, 3).
Value
A named list of results, one element per binary outcome variable. Each element is a list containing:
- table
A tibble with summary statistics and model predictions per level of `caika` and `cever`.
- plot1
A ggplot object of the log-scale SIR across `caika` or `cever`.
- plot2
A ggplot object showing adjusted exposure by Age and response status.
Details
For each binary outcome, two Poisson regression models are fitted:
One using `caika` (calendar period) as the main time variable
One using `cever` (time since event or similar)
Both models adjust for age using restricted cubic splines and use log(pyrs) as offset. Standard errors are heteroskedasticity-consistent (HC0). Predictions are made at Age = 70 and a base rate of 10,000 person-years.