R Project 3

Thesis: In West Virginia, the strongest predictors of county-level educational outcomes are average family income and the percentage of residents with at least a bachelor’s degree.

This analysis examines various data points and factors to predict educational outcomes at the county level in West Virginia. The predictors tested for their influence on proficiency rates include federal, state, and local revenues allocated to each county; county-level expenditures; unemployment rates; the percentage of residents with less than a 9th-grade education; the enrollment amount per each county; the percentage with less than a high school education; the percentage with at least a bachelor’s degree; and income metrics such as average vast majority income, average household income, and average family income. These factors are analyzed to determine their impact on past proficiency rates and identify the strongest influences on educational outcomes.

Data Description

The data was gathered from us census websites for West Virginia counties.

Key variables included:

• enroll: The amount of students enrolled in each county.
• tfedrev: The amount of federal revenue allocated to each county.
• tstrev: The amount of state revenue allocated to each county.
• tlocrev: The amount of local revenue allocated to each county.
• totalexp: The amount of expenditures utilized by each county.
• unemployed: The percent of unemployment per each county.
• less_than_9th_grade_education: The percent of people in each county with less than a 9th grade education.
• less_than_high_school_grade_education: The percent of people in each county with less than a high school education.
• at_least_bachelor_education: The percent of people in each county with at least bachelor’s degree.
• vast_majority_income: The average amount of vast majority income per each county.
• household_income: The average amount of household income per each county.
• family_income: The average amount of family income per each county.
• proficiency: The percent of students that are educationally proficient per each county.
• proficiency_range: A range of percents of students that are educationally proficient per each county.

Methods

Load Assessment Data

In this section, the assessment data provided by the professor was loaded.

Load Spending Data

In this section, the spending data provided by the professor was loaded.

Load Demographic Data

In this section, the demographic data provided by the professor was loaded.

Add in New Data

In this section, new data was loaded and cleaned. Additional educational outcome data, including math proficiency, reading proficiency, and an average proficiency score, was incorporated. County-level educational attainment data and income level information were also added.

Join Data

In this section, the loaded data was merged to create a unified dataset, facilitating more efficient analysis and ensuring all relevant variables are accessible for the study.

View Proficiency Data

In this section, a state-level graph was created to visualize proficiency rates at the county level.

Correlations

In this section, correlation analysis was performed to identify the factors with the greatest influence on educational proficiency.

Create Test/Training Data

In this section, the data was divided into training and testing sets to facilitate model training and evaluation.

Linear Regression Model

In this section, a linear regression model was created to predict educational proficiency levels.

Call:
lm(formula = proficiency ~ at_least_bachelor_education + family_income, 
    data = t_train)

Residuals:
    Min      1Q  Median      3Q     Max 
-6.4377 -2.0840 -0.2013  1.9891  8.5273 

Coefficients:
                               Estimate  Std. Error t value Pr(>|t|)   
(Intercept)                 12.01459814  3.64431128   3.297  0.00200 **
at_least_bachelor_education  0.32319107  0.11888552   2.719  0.00949 **
family_income                0.00017861  0.00007458   2.395  0.02117 * 
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 3.776 on 42 degrees of freedom
Multiple R-squared:  0.5767,    Adjusted R-squared:  0.5566 
F-statistic: 28.61 on 2 and 42 DF,  p-value: 0.00000001442

View Residuals of Linear Regression Model

In this section, a histogram was created to visualize the distribution of residuals based on the linear regression model.

PCA

In this section, principal component analysis (PCA) was performed to identify key patterns in the population.

K-Means

In this section, k-means clustering was performed to group the population into distinct clusters based on similar patterns.

Decision Tree

In this section, a decision tree was created to classify the population based on distinct patterns and key factors.

Decision Tree Results

In this section, a confusion matrix was generated to evaluate the performance of the decision tree by comparing the predicted values with the actual outcomes.

[1] 0.6363636
          
           18 to 25 25 to 30 30 to 35 35 to 40 40 to 47
  18 TO 25        4        0        3        0        0
  25 TO 30        5       22        3        2        1
  30 TO 35        0        2        5        1        0
  35 TO 40        0        0        1        4        2
          
             18 to 25   25 to 30   30 to 35   35 to 40   40 to 47
  18 TO 25 0.44444444 0.00000000 0.25000000 0.00000000 0.00000000
  25 TO 30 0.55555556 0.91666667 0.25000000 0.28571429 0.33333333
  30 TO 35 0.00000000 0.08333333 0.41666667 0.14285714 0.00000000
  35 TO 40 0.00000000 0.00000000 0.08333333 0.57142857 0.66666667

Limitations

The timing of the data was inconsistent. My demographic, education, and income data are based on the most recent information, while the educational proficiency rate data is a few years older. As a result, the county-level data may have been different a few years ago, which could have affected the outcomes. Additionally, there are several factors I didn’t consider, such as data on incarceration, smoking rates, non-English speaking populations, and other variables that could influence proficiency rates.

References

Sources included:

• West Virginia County-Level Income Data: https://hdpulse.nimhd.nih.gov/data-portal/social/map?socialtopic=030&socialtopic_options=social_6&demo=00010&demo_options=income_3&race=00&race_options=race_7&sex=0&sex_options=sexboth_1&age=001&age_options=ageall_1&statefips=54&statefips_options=area_states
• West Virginia County-Level Educational Data: https://hdpulse.nimhd.nih.gov/data-portal/social/map?socialtopic=020&socialtopic_options=social_6&demo=00006&demo_options=education_3&race=00&race_options=race_7&sex=0&sex_options=sex_3&age=081&age_options=age25_1&statefips=54&statefips_options=area_states
• ChatGPT: https://openai.com/chatgpt/overview/ ChatGPT helped me throughout the project. I would write the code and ask ChatGPT for help if I ran into problems or wasn’t sure how to do something. I also asked ChatGPT for advice on how to work and achieve my output efficiently.