---
title: "Examples: Centered Wald Latency"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Examples: Centered Wald Latency}
  %\VignetteEngine{knitr::rmarkdown}
  \usepackage[utf8]{inputenc}
---

```{r setup}
source("lkt-vignette-setup.R")
```

The Wald model uses one normal trial-noise draw and a fixed participant speed effect.
The Wald contamination allowance is used for fitting only, never generation.

```{r fit-latencies}
set.seed(20261030)
sid <- rep(paste0("s",1:6),each=16)
d <- data.table::data.table(Anon.Student.Id=sid,
  KC..Default.=rep(c("k1","k2"),48),
  CF..ansbin.=rep(c(1,1,0,1,0,0,1,1),12),
  CF..Time.=rep(seq(1000,1150,10),6),
  Duration..sec.=rep(seq(3,25,length.out=16),6),fold=1L)
wald <- LKT(data.table::copy(d),components="KC..Default.",features="logitdec",
  fixedpars=.9,interc=TRUE,dualfit=TRUE,distribution="Wald",verbose=FALSE,
  wald_t0_fixed=0,wald_control=list(nodes=15L,maxit=120L))
check_true("Wald result",wald$model_specification[[2]]$distribution=="Wald")
check_true("Wald engaged generation",wald$model_specification[[2]]$generation=="engaged_only")
check_true("Wald finite prediction",all(is.finite(wald$latencymodel[[1]]$predict(c(.2,.5,.9),0,FALSE))))
```

The `usefolds` and optional `latency_question_type_filter` settings select fitting
observations. Excluded rows still contribute to the actual feature histories.
The retired exponential and Weibull latency choices are deliberately rejected.