What is a Better Predictor of Number of Deaths by Cancer:

Smoking Rate or Obesity Rate?

Priscilla Maccario

Introduction

My study consists of data taken in 2013, across all fifty states. Percent of adults smoking per state and percent of obese adults per state are the explanatory variables. The response variable is the rate of cancer death per 100,000 people for each state. I predict that obesity rate will have a higher, positive correlation with number of deaths by cancer than smoking rate. My reason is that I think there are more obesity related cancers than smoke related cancers.

Data Collection Pt. 1

I obtained the data from websites cited on the next slide.

Data Collection Pt. 2

Bach, Laura. “Key State-Specific Tobacco-Related Data & Rankings” 22 July 2015. Web. 7 Oct. 2015

​

"Adult Obesity in the United States." : The State of Obesity. N.p., n.d. Web. 7 Oct. 2015.

​

"Cancer Death Rate per 100,000 Statistics - States Compared - State Master." RSS Stats. N.p., n.d. Web. 9 Oct. 2015.

Graphs

The scatterplots show that there is a higher prevalence of obesity rate in states than smoking rate. However, both show similar rates of death by cancer. There is one outlier for both graphs and may be the same state.

Percent Adult Obesity Rate

Numerical Summaries and Interpretations

Obesity rate and cancer deaths had a slightly higher correlation than smoking rate and cancer deaths. The obesity and cancer correlation starts off at a lower y-intercept than smoking and increases more for each death by cancer. The r-squared value for obesity reveals that 34.5% of the variability of cancer deaths can be explained using the linear model of adult obesity rate. For the smoking r-squared value, 31.3% of the variability of cancer deaths can be explained using a linear model of smoking rate.

Conclusion and Discussion

Obesity rate does a slightly better job in predicting number of deaths by cancer. It had a higher correlation and a higher r-squared value, so it could explain more of the variability in the data using a linear model than could smoking rate. The outlier in both data sets is Alaska. This may be because of the cleaner air and other variables. It would be appropriate to use the least-squares regression line for smoking rate between 10-30%. It would be most appropriate to use the least-squares regression line for obesity rate between 20-35%.