Skip to content
This repository has been archived by the owner on May 14, 2024. It is now read-only.

A semester research project for my STAT2332 class to store my methods/findings.

License

Notifications You must be signed in to change notification settings

JonathanLTurner03/STAT2332-Degree_Unemployment_Research

Repository files navigation

STATS2332 - Probability and Data Analysis - Final Project

Background & Dataset

This project was completed in RStudio and was conducted over a dataset provided by fivethirtyeight posted on data.world. This project aims to test for a statistical signficance between a degree's college (ie. Engineering, Business, etc.) and the unemployment rate of the sampled individuals. The dataset contains information on 173 college majors and 173 rows of data. The dataset contains many different variables. However, in this project the data is slimmed to only two columns: Major_category and Unemployment_rate.

Research Question

Aiming the direction for this project, I aimed to answer the following research question:

How much does a graduate student’s college affect the Return on Investment (ROI) a student may receive in regard to the market’s unemployment percentage?

From this research question, I created two hypotheses to test for a statistical significance between the two variables.

Null Hypothesis

A graduate student’s college does NOT affect the ROI a student may receive in regard to the market’s unemployment percentage.

Alternative Hypothesis

A graduate student’s college DOES affect the ROI a student may receive in regard to the market’s unemployment percentage.

Libraries Used

Library Version
ggplot2 3.5.1
tidyverse 2.0.0

Data Filtering & Trimming

Trimming the Data

The dataset contains many different variables. However, in this project the data is slimmed to only two columns: Major_category and Unemployment_rate. The data is trimmed to only these two columns to focus on the research question and to make the data easier to work with. This is done by creating a new data.frame() with only the two columns. These columns were remapped to college: dataset$Major_category and rate: dataset$Unemployment_rate.

Dataset Prior to Trimming:

Dataset Post Trimming:

Removing NA Values

The data was also cleaned by removing any NA values. This was done by using the na.omit() function in RStudio. This function removes any rows with NA values. This was done to ensure that the data was clean and that the data was not skewed by any NA values. This was specifically done post trimming to prevent any NA values from other, unneeded columns, from altering the data we were working with. This resulted in only 1 row being omitted from our dataset besides 15.

Methodology

ANOVA

To test for a statistical significance, I used an ANOVA test to compare the means of the unemployment rates of the different college majors. The ANOVA test was chosen because it is used to compare the means of three or more groups, in which I had over 16. I conducted these tests in RStudio and used the aov() function.

Outlier Detection and Impact

I also conducted an outlier detection test to see if any of the data points were skewing the results. I used the boxplot() function in RStudio to visualize the data and see if any outliers were present. There were some outliers on present in the data. However, I conducted the ANOVA test with and without the outliers to see if the outliers were skewing the results.

This is the result of the boxplot() function:

Results

ANOVA Test

The ANOVA test was conducted with and without the outliers. The results of the ANOVA test are as follows:

Test p-value Result
With Outliers 0.00101 Reject Null Hypothesis
Without Outliers 0.000209 Reject Null Hypothesis

Did the Outliers Impact the Results?

The results of the ANOVA test show that the outliers did impact the results, however the results of the ANOVA test with and without the outliers were the same. This means that the outliers did not skew the results of the ANOVA test.

Conclusion

Results

The results of the ANOVA test show that the null hypothesis can be rejected. This means that a graduate student's college does affect the ROI a student may receive in regard to the market's unemployment percentage.

Data Representation

While the data of the results are significant, the data is not conclusive on which majors are the most affected by the market's unemployment percentage. While each degree has its own unemployment rate, the data does not show which degrees are the most affected by the market's unemployment rate. If the data is represented as the mean unemployment rate of each college, the data levels out and the data is not as significant. However, the data does show that the null hypothesis can be rejected and that a graduate student's college does affect the ROI a student may receive in regard to the market's unemployment percentage.

Bar Graphs of Data

Data represented by each major inside a college vs the mean of each college:

About

A semester research project for my STAT2332 class to store my methods/findings.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages