CDA-LogLinearModel.Rmd

---
title: "Loglinear Model Analysis"
author: "Riley Smith"
date: "`r format(Sys.Date(), '%d %B %Y')`"
---

```{r setup, echo=FALSE, message=FALSE, warning=FALSE, cache=FALSE}
source("../SETUP.R")
options(width = 70)
knitr::opts_chunk$set(
    tidy = FALSE,
    echo = TRUE,
    cache = FALSE,
    fig.keep = 'high',
    fig.show = 'asis',
    results = 'asis',
    autodep = T,
    Rplot = TRUE,
    dev = 'pdf',
    fig.path = 'graphics/reuters/rplot_',
    fig.width = 7,
    fig.asp = 1,
    out.width = "\\linewidth"
    )
library(foreign) ## read.spss() ##
library(car) ## recode() ##
```

`r tufte::margin_note("Data: \\textit{\\color{dkpatriarch}{\\texttt{data/reuters.sav}}}")`

$$ \log(\mu_{ij}) = \lambda + \lambda_{i}^{X} + \lambda_{j}^{Y} + \lambda_{ij}^{XY} $$

```{r}
dat <- R.rspss("data/reuters.sav", vlabs = F)
R.msmm(dat)
dat <- dat[, -3]
dat <- subset(dat, response < 2)
dat <- within(dat, {
    ind.f <- factor(ind,
                    levels = unique(ind),
                    labels = c("Independent",
                               "Affililate"))
    response.f <- factor(response,
                        levels = c(0, 1),
                        labels = c("Trump",
                                   "Clinton"))
})

tbl <- table(dat$ind.f, dat$response.f)
dimnames(tbl) <- list(ind = levels(dat$ind.f),
                      response = levels(dat$response.f))
mosaicplot(tbl, type = "deviance", las = 2, color = pal_my.a75[c(5, 16)])

library(MASS)
logmodel <- loglm( ~ ind + response, digits = 4, data = tbl)

```

\newpage
# References`r R.cite_r("~/GitHub/auxDocs/REFs.bib", footnote = TRUE)`