https://github.com/kklot/ordinaltmb

https://github.com/kklot/ordinaltmb

Last synced: 12 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/kklot/ordinaltmb
• Owner: kklot
• License: other
• Created: 2023-08-27T11:10:06.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-09-14T18:03:06.000Z (over 1 year ago)
• Last Synced: 2023-09-15T10:57:21.346Z (over 1 year ago)
• Language: R
• Homepage:
• Size: 44.9 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.Rmd
  • License: LICENSE

Awesome Lists containing this project

README

        
---

output: github_document

---

```{r, include = FALSE}

knitr::opts_chunk$set(

  collapse = TRUE,

  comment = "#>",

  fig.path = "man/figures/README-",

  out.width = "100%"

)

```

# ordinalTMB - univariate and multivariate ordinal regresison

[![Lifecycle: experimental](https://img.shields.io/badge/lifecycle-experimental-orange.svg)](https://lifecycle.r-lib.org/articles/stages.html#experimental)

This package is written for the "Gender Across Spectrum" paper (doi to be

updated), expecting to be reused further for Likert scales outcomes studies.

## Why?

Run an ordinal model with random effects, but:

- `ordinal::clmm` sometimes gave all the parameter estimates the same SE, not sure why; nevertheless, the ultimate aim is to do Bayesian inference

- `rstanarm::stan_polr` does not provide random effect specification (?)

- `brms::brm()` with `cummulative` family seems like the one, but failed to compile on MacOS, or an HPC server (scientific linux), and on both systems with a separated `miniconda` environment.

  - on both systems, other packages using Eigen, Armadillo, Boost, INLA,...work fine.

  - the data is large, so `brm` would be slow anyway

- `INLA` [does not have ordinal model since 2015?](https://groups.google.com/g/r-inla-discussion-group/c/7Tl6DanHmtM/m/agDFK0w23VAJ)

Moreover, to do multivariates ordinal regression

- similar to `mvord`, `PLordprob`, which do not have random effect specification

## TODO

By priorities

- [x] Fit a simple ordinal with logistic link and compare with `clm` estimates

- [x] Fit an ordinal logistic with random intercepts and compared with `clmm`

- [x] add probit link and compare with `clmm`, `clm`

- [ ] add multivariates ordinal

- [ ] add other links (c-log-log,...)

- [ ] add glm-ism formula interface?

## Temporarily workflow - univariate ordinal

1. Clone the package

1. Prepare data and meta parameters for TMB like usual, example:

```r

data <- full_data |> na.omit()

kdata = list()

kdata$y <- data$outcome # coded as 1, 2,... as its order

fm <- y ~ -1 + q + time + age # no intercept and with covariates

# Fix effect

mf <- model.frame(fm, data)

kdata$X <- model.matrix(fm, mf)

(J <- max(kdata$y)) # number of categories

(K <- ncol(kdata$X)) # number of covariates

# random effect

kdata$R <- make_iid_matrix(data$id)

(S <- length(unique(data$id)))

# TMB parameters

kpar <- list(

  pi_norm = rnorm(J - 1),

  beta = rnorm(K),

  iid = rnorm(S), 

  sd_log = log(1)

)

```

2. Edit the TMB model files in `ordinaTMB/src/ordinalTMB.cpp` according to your need

```r

ktools::tmb_unload('ordinalTMB'); # run this every time you edit the model file

devtools::load_all("~/ordinalTMB/") # load the package

model <- ktools::MakeADFunSafe( # avoid crashing R's session

    data = kdata,

    parameters = kpar,

    random = c('iid'),

    DLL = "ordinalTMB"

)

fit <- nlminb(model$par, model$fn, model$gr)

model$report() |> glimpse()

```