Multinomial Regression.Rmd

---
title: "IMMERSE_Day 5_Multinomial Regression"
author: "Netasha"
date: "2024-05-20"
output: html_document
---

1. From github select github repository
2. Click the "<> Code" 
3. Click Download Zip
4. Move Zip file to desktop
5. Make sure file is unzipped!
6. Open IMMERSE_Day-5.Rproj project
7. Select the "Multinomial regression.Rmd"

## Read in packages
There are a few packages we want our markdown to read in:
Click on the "run current chunk" button (looks like a play button on line 19 )

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(MplusAutomation) #use this code to run mplus code in R
library(tidyverse) #collection of R packages designed for data science
library(here) #helps with filepaths
library(haven) # read_sav()
library(psych) # describe()
library(ggpubr) # ggdensity() and ggqqplot()
library(corrplot) # corrplot()
here::i_am("Multinomial Regression.Rmd")
```

## Overview: Camera Marketing Study
Sample of 735 of individuals surveyed by a market research group for the purposes of investigating the role of age and “gender” (outdated binary—this is old data) in digital camera brand choices.

Variables for the study include:

1.  Brand - 1 = Cannon - 2 = Kodak - 3 = Nikon
2.  Female - 1 = Female - 0 = Male
3.  Age (in years)

## Import Data

 In R:
-   Start a new code chunk to load the data
    -   PC: Ctrl + Alt + I
    -   Mac: Option + Command + I

```{r}
#Using the read_csv code we will import our .dat data file into R. We are going to add column names to descriptive purposes
#there are many ways to import your data such as the import dataset menu in the environment (point and click approach)
data_csv <- read_csv(here("camera.dat"), col_names = c("brand", "female", "age"))
```

## Evaluate Data 
Create a new code chunk
Use the summary function to get summary of the data 
Lets make a contingency (3x2) table to get a snapshot of camera choice using the table function
-   Start a new code chunk to load the data
    -   PC: Ctrl + Alt + I
    -   Mac: Option + Command + I
- use the summary function 
- use the table function 
  - create a name for the table
  - Assign the table using assignment operator 
    -   PC: Alt + (-)
    -   Mac: Option + (-) 
  - type table
  - type the name of the data set 
  - type $
  -use the drop down menu to locate the variable or begin to type the variable name "brand" (x-axis) and select
  - add a comma to add the second variable 
  - type the name of the data set followed by the dollar sign and "female" (y-axis)
  - note: the table function will give a basic table. Labels will not be presented on the x and y axis. We would need to change the variables to factors.
  

```{r}
summary(data_csv)
table1 <-table(data_csv$brand, data_csv$female)
table1
```
## Prepare data for mplus 
Introduced last week: 

Convert from file to Mplus *.dat* file**

Say you want an Mplus .dat file and don't want to go through the hassle of deleting rows and manual conversion to *.dat*.
You can use the `prepareMplusData()` function to convert from *.csv* to *.dat*. A new file will be generated in your folder 

create a new code chunk
```{r}
prepareMplusData(data_csv, here("exp_data.dat"))
```
## Using `MplusAutomation` for multinomial regression 

Reminder from last week:
**What does the** `mplusObject()` **function do?**
1. It generates an Mplus input file (does not need full variable name list, its automated for you!) 
2. It generates a datafile specific to each model 
3. It runs or estimates the model (hopefully) producing the correct output. Always check!

**What does the** `mplusModeler()` **function do?**
1.  Creates, runs, and reads Mplus models created using `mplusObject()`
2.  You can specify where you want the *.out* file saved
3.  `check=TRUE` checks for missing semicolons, `run=TRUE` runs the model, `hashfilename=FALSE` does not add a hash of the raw data to the data file name.

create a new code chunk 

### Mulinomial regression of brand on female simple 

```{r, echo=TRUE, eval=FALSE}
m_basic_simp  <- mplusObject(
  
  TITLE = "Multinomial regression brand on female;",
  
  VARIABLE = 
 "usevar = brand female;
 Nominal are brand;",
  
  ANALYSIS = 
 "estimator = ML;",
 
 MODEL = 
   "brand on female;",
 
  usevariables = colnames(data_csv), 
  rdata = data_csv)

m_basic_fit_simp <- mplusModeler(m_basic_simp, 
               dataout=here("basic_simp.dat"),
               modelout=here("basic_simp.inp"),
               check=TRUE, run = TRUE, hashfilename = FALSE)
```
### Multiple regression with MLR and selected output 
```{r, echo=TRUE, eval=FALSE}
m_basic  <- mplusObject(
  
  TITLE = "Multinomial regression brand on female;",
  
  VARIABLE = 
 "usevar = brand female;
 Nominal are brand;",
  
  ANALYSIS = 
 "estimator = MLR;",
 
 MODEL = 
   "brand on female;",
 
 OUTPUT = 
   "sampstat crosstabs residual cinterval tech1 svalues;",
 
  usevariables = colnames(data_csv), 
  rdata = data_csv)

m_basic_fit <- mplusModeler(m_basic, 
               dataout=here("basic.dat"),
               modelout=here("basic.inp"),
               check=TRUE, run = TRUE, hashfilename = FALSE)
```

CHECK YOUR OUTPUT!!! Select .out to view output!

### Use final estimates as starting values in the command line 

create new code chunk
-Mulinomial regression of brand on female with starting values from last output. 
- must add a * before start value 
-Add parameter labels to each level of brand in parenthesis 

```{r, echo=TRUE, eval=FALSE}
m_basic_2  <- mplusObject(
  
  TITLE = "Multinomial regression brand on female;",
  
  VARIABLE = 
 "usevar = brand female;
 Nominal are brand;",
  
  ANALYSIS = 
 "estimator = MLR;",
 
 MODEL = 
   "brand#1 on female* -0.38299 (femCvN);
    brand#2 on female* -0.13628 (femKvN);
    [brand#1*0.16508] (intCvN);
    [brand#2*0.23841] (intKvN);",
 
 MODELTEST = 
     "0 = femCvN;
    0 = femKvN;",
 
  usevariables = colnames(data_csv), 
  rdata = data_csv)

m_basic_fit_2 <- mplusModeler(m_basic_2, 
               dataout=here("basic_2.dat"),
               modelout=here("basic_2.inp"),
               check=TRUE, run = TRUE, hashfilename = FALSE)
```

### Mplus input with alternate COR/RR
```{r, echo=TRUE, eval=FALSE}
m_basic_3  <- mplusObject(
  
  TITLE = "Multinomial regression brand on female;",
  
  VARIABLE = 
 "usevar = brand female;
 Nominal are brand;",
  
  ANALYSIS = 
 "estimator = MLR;",
 
 MODEL = 
   "brand#1 on female* -0.38299 (femCvN);
    brand#2 on female* -0.13628 (femKvN);
    [brand#1*0.16508] (intCvN);
    [brand#2*0.23841] (intKvN);",
 
 MODELCONSTRAINT = 
   "New(femCvK efemCvK);
       femCvK = femCvN-femKvN;
       efemCvK = exp(femCvk);",
 
  usevariables = colnames(data_csv), 
  rdata = data_csv)

m_basic_fit_3 <- mplusModeler(m_basic_3, 
               dataout=here("basic_3.dat"),
               modelout=here("basic_3.inp"),
               check=TRUE, run = TRUE, hashfilename = FALSE)
```

### Mplus input multinomial regression of brand on female & age
```{r, echo=TRUE, eval=FALSE}
m_basic_4  <- mplusObject(
  
  TITLE = "Multinomial regression brand on female and age;",
  
  VARIABLE = 
 "usevar = brand female age;
 Nominal are brand;",
  
  ANALYSIS = 
 "estimator = MLR;",
 
 MODEL = 
   "brand on female age",
 
OUTPUT = 
  "sampstat crosstabs residual cinterval tech1 svalues;",
 
  usevariables = colnames(data_csv), 
  rdata = data_csv)

m_basic_fit_4 <- mplusModeler(m_basic_4, 
               dataout=here("basic_4.dat"),
               modelout=here("basic_4.inp"),
               check=TRUE, run = TRUE, hashfilename = FALSE)
```