Alzheimer’s Trial: Linear Progression with Delayed Treatment Onset

This vignette analyzes a simulated ADNI-style trial using pmrm_delayed() with profile-likelihood delay estimation.

library(progme)
data(adni_simulated)

fit <- tryCatch(
  pmrm_delayed(adas_cog ~ time * trt, data = adni_simulated,
               subject_var = "rid",
               delay_estimation = "profile",
               delay_grid = seq(0, 18, by = 3), lambda = 0.5),
  error = function(e) e
)
if (!inherits(fit, "error")) {
  print(fit)
  s <- summary(fit)
  head(s$t_table)
} else {
  message("Demo fit skipped: ", conditionMessage(fit))
}
#> progme delayed-effect progression model
#> Progression shape : linear 
#> Estimated delay   : 6 
#> Delay CI          : [6, 6]
#>                  Value  Std.Error   DF   t-value       p-value
#> (Intercept) 22.2602437 0.12091625 1883 184.09637  0.000000e+00
#> time         1.2254149 0.03133736 1883  39.10396 2.331909e-245
#> trt_eff     -0.5544902 0.03120106 1883 -17.77152  1.919385e-65

The profile log-likelihood and its chi-squared threshold give an approximate confidence interval for the onset delay.