Password caching

If you would like your git password cached for a week for this project, type the following in the Terminal:

git config --global credential.helper 'cache --timeout 604800'

You will need to enter your GitHub username and password one more time after caching the password. After that you won’t need to enter your credentials for 604800 seconds = 7 days.

Packages

You will need the following packages for today’s lab:

library(tidyverse)
library(nnet)
library(knitr)
library(broom)
# Fill in other packages as needed

Data

The data for this lab are from the 2016 General Social Survey. The original data set contains 2867 observations and 935 variables. Given the size of the dataset, we will handle it differently in our workflow than we’ve handled data in previous assignments.

⊕Working with large files

The size of this dataset is 34.3 MB. Compare that to the Spotify dataset from last weeks’ lab which was 149 KB (0.149 MB)! GitHub will not allow you to push files larger than 100 MB and will give you a warning when you push files as large as 50 MB. Though we could push the file we’re working with today to GitHub, it’s large enough that we’d still prefer not to.

You have may noticed that each repo contains a file called .gitignore. It contains a list of the files you don’t want commit or push to GitHub. If you look at the .gitignore file for today’s lab, you will notice that gss2016.csv is listed at the bottom.

• Click here to download gss2016.csv.
• Upload gss2016.csv into the data folder of your project.
• Notice that gss2016.csv does not appear in your Git pane. This is because it is being ignored by git, since it is listed in the .gitignore file.

You will use the following variables in the lab:

• natmass: Respondent’s answer to the following prompt:

  “We are faced with many problems in this country, none of which can be solved easily or inexpensively. I’m going to name some of these problems, and for each one I’d like you to tell me whether you think we’re spending too much money on it, too little money, or about the right amount…are we spending too much, too little, or about the right amount on mass transportation?”

• age: Age in years.

• sex: Sex recorded as male or female

• sei10: Socioeconomic index from 0 to 100

• region: Region where interview took place

• polviews: Respondent’s answer to the following prompt:

  “We hear a lot of talk these days about liberals and conservatives. I’m going to show you a seven-point scale on which the political views that people might hold are arranged from extremely liberal - point 1 - to extremely conservative - point 7. Where would you place yourself on this scale?”

Use the code below to read in the data.

gss <- read_csv("data/gss2016.csv",
  na = c("", "Don't know", "No answer", 
         "Not applicable"), 
         guess_max = 2867) %>%
  select(natmass, age, sex, sei10, region, polviews) %>%
  drop_na()

The argument guess_max = 2867 tells the read_csv function to use all of the observations in a column to determine its data type. Without this argument, only the first 1,000 observations would be used to make this determination. This becomes important for a variable like age; though age is coded as numeric data for most of the observations, there are some in which age is coded as "89 or older". Without the guess_max argument, you will get warnings when loading the data.

Note also that only the variables of interest will be loaded, not the entire dataset. This will make for faster computation and knitting as you work on the lab.

Part I: Exploratory Data Analysis

⊕See Reorder factor levels by hand for documentation about fct_relevel.

1. The variable natmass will be the response variable in the model, and you want to compare more opinionated views to the moderate position. Recode natmass so it is a factor variable with "About right" as the baseline.

2. Recode polviews so it is a factor variable type with levels that are in an order that is consistent with question on the survey. Note how the categories are spelled in the data.

   Make a plot of the distribution of polviews. Which political view occurs most frequently in this data set?

3. Make a plot displaying the relationship between natmass and polviews. Use the plot to describe the relationship between a person’s political views and their views on mass transportation spending.

4. You want to use age as a quantitative variable in your model; however, it is currently a character data type because some observations are coded as "89 or older". Recode age so that is a numeric variable. Note: Before making the variable numeric, you will need to replace the values "89 or older" with a single value.

Part II: Multinomial Logistic Regression Model

5. You plan to fit a model using age, sex, sei10, and region to understand variation in opinions about spending on mass transportation. Briefly explain why you should fit a multinomial logistic model.

6. Fit the model described in the previous exercise and display the model output. Make any necessary adjustments to the variables so the intercept will have a meaningful interpretation. Be sure About Right is the baseline level. Be sure the full model displays in the knitted document.

7. Interpret the intercept associated with odds of having an opinion of “Too much” versus “About right”.

8. Consider the relationship between age and one’s opinion about spending on mass transportation. Interpret the coefficient of age in terms of the odds of having an opinion of “Too little” versus “About right”.

9. Now that you have adjusted for some demographic factors, let’s examine whether a person’s political views has a significant impact on their attitude towards spending on mass transportation.

   Conduct the appropriate test to determine if polviews is a significant predictor of attitude towards spending on mass transportation. State the null and alternative hypothesis, display all relevant code and output, and state your conclusion in the context of the problem.

   Choose the appropriate model based on the results from the test. Use this model for the next part of the lab.

Part III: Model Fit

11. Calculate the predicted probabilities and residuals from your model.

12. Let’s make some of the plots and tables you use to check the linearity assumption for multinomial logistic regression. Plot the binned residuals versus the predicted probabilities for each category of natmass. You will have three plots.

⊕You can change the size of your plots, so you can fit multiple plots on a single page. Include the arguments fig.height = and fig.width = in the header of the code chunk to change the plot size. See Using R Markdown for an example.

13. To examine the residuals versus each categorical predictor, you will look at the average residuals for each each category of the categorical variables.

    • For each category of natmass, calculate the average residuals across categories of region.

Based on the plot and table above, discuss with your group whether there are any obvious violations of the linearity assumption. Note that we haven’t examined all of the plots and tables of the residuals needed to make an assessment about the linearity assumption.

The other assumptions are randomness and independence. Discuss with your group whether these assumptions are satisfied for this analysis.

Part IV: Using the Model

16. Use your model to describe the relationship between one’s political views and their attitude towards spending on mass transportation.

17. Use your model to predict the category of natmass for each observation in your dataset. Display a table of the actual versus the predicted natmass. What is the misclassification rate?

Acknowledgements

The “Data” section is largely inspired by datasciencebox.org.