Correlation matrix:

These results did not show what I expected every time. Our first independent variable was the uninsured rate. It has a moderate negative correlation with per pupil spending of 0.39. This means that there is a correlation of states with more uninsured people will spend less on education. This is mostly what I expected, healthcare would likely take priority over education. Having many people without insurance likely takes up many government resources. Could also be a correlation that is tied to income, people with low incomes are less likely to have insurance and those states would then have less tax revenue leading to budget cuts. Could also be a correlation between states that decided not to expand medicaid (leading to more people to be uninsured) and states with republican controlled legislatures and govenorships which often support cutting funding for public education in order to increase funding for charter schools or other private schools.
The second independent variable is the police expenditure per state. This did not meet my expectations. I expected states with higher police expenditures to have lower per pupil spending since there are limited resources within state governments. If you spend more on education, I thought it would mean less spending on police. However, there is a moderate positive correlation between state police expenditure and per pupil spending. This is not one I have many theories on other than states with high revenues tend to spend more on both police and K12 education.
The third independent variable I used was the seats in state legislatures held by women. Somewhat stereotypically of me, I expected a moderate to high correlation between this two. I thought that women in state legislatures would want more education spending and women (especially in state legislatures) tend to be more liberal. What I found was a weaker correlation that I expected. The correlation was a weak to moderate positive correlation of 0.25. Even then, the pvalue is not outside of where most tests would put the cut off. I would not reject the null hypothesis that there is no correlation.
Multiple linear regression results:
Dependent Variable: Per Pupil Spending, Public K12 ($) Independent Variable(s): Uninsured (% of individuals lacking coverage), State Police Expenditure ($ per resident), Seats in State Legislatures Held by Women (%) Per Pupil Spending, Public K12 ($) = 9901.5985 + 174.44187 Uninsured (% of individuals lacking coverage) + 38.795043 State Police Expenditure ($ per resident) + 47.310693 Seats in State Legislatures Held by Women (%) Parameter estimates:
Analysis of variance table for multiple regression model:
Summary of fit: Root MSE: 1940.7161 Rsquared: 0.3529 Rsquared (adjusted): 0.3107 
The direction of the relationships are the same as above. The uninsured rate has a negative correlation while police expenditure and seats held by women has a positive correlation.
The uninsured rate and state police expenditure seems to be moderate to significant; however, the pvalue for seats held by women is too high for us to comfortably reject the null hypothesis. The correlation may not be strong enough for any conclusions to be made.
The Tstat for police expenditure is the highest meaning that it has the greatest correlation with our dependent variable. However, there is greater perunit change for % uninsured. This is because we are comparing different units when talking about dollars spent versus percentage points.
Our rsquared value is 0.3529, meaning the model explains around 35% of the variability in per pupil spending. This isn't exactly a great fit, but there certainly is some correlation amongst these variables.
I would say that this explains less I originally expected from these variables. I wanted to go with something outside of the box that would not be the most obvious to choose. However, there was still a fairly strong correlation between uninsured percentage/state police expenditure and per pupil spending. For every percentage point increase in the uninsured rate, we could expect to see $174 less in per pupil spending. For ever extra dollar spend on the police force, we could expect to see an extra $39 in per student spending. I am sure that there are plenty more variables that could be correlated with education spending and it would be very interesting to analyze those as well.
Already a member? Sign in.