Principal Component Analysis
PCA is conducted to select two more most correlated features from moon illumination, max temperature, min temperature, day, month, week day/weekend
Comparing the absolute value of each components, the two most important components are moon illumination and max temperature.
Hypothesis Testing
Rainy Weather vs. Crime type:
H0: Crime rates of varied types are not affected by rainy weather.
HA: The crime rates are higher when it’s rainy than not rainy.
The confidence level is set to 0.95.
For Assault, Death, Drug, Fraud and Robbery, the crime rates are not affected by rainy weather. However, the crime rates of Burglary, Sexual and Theft is lower when it’s rainy.
New Moon vs. Crime type:
H0: Crime rates of varied types are not affected no matter what the moon phases is.
HA: The crime rates are higher when it’s new moon.
The confidence level is set to 0.90.
From the table above, only the crime rates of Assault, Robbery are affected by New Moon and are higher than usual, other crime types are unaffected. Compared to the line plot above, though they appeared to be affected by New Moon, in fact Burglary, Drug, Sexual and Theft are proved to not have statistical significance.
Chi-square Test of Independence
Education
H0: Education level is independent of crime type (there is no association between education level and crime).
HA: Education level is not independent of crime type (there is an association between education level and crime).
From the above results, only Drug and Robbery crime have p-value below 0.05, which means there is enough evidence to reject the null hypothesis, and accept the alternative hypothesis. Thus, education level is not independent with Drug and Robbery.
Health (Obesity)
H0: Adult obesity rate is independent of crime type (there is no association between obesity rate and crime).
HA: Adult obesity rate is not independent of crime type (there is an association between obesity rate and crime).
From the above results, all the crime types have a p-value below 0.05, which means there is enough evidence to reject the null hypothesis, and accept the alternative hypothesis. Thus, adult obesity rate is dependent with different crime types.
Health (Smoking Rate)
H0: Adult smoking rate is independent of crime type (there is no association between smoking rate and crime).
HA: Adult smoking rate is not independent of crime type (there is an association between smoking rate and crime).
From the above results, all the crime types have a p-value below 0.05, which means there is enough evidence to reject the null hypothesis, and accept the alternative hypothesis. Thus, adult smoking rate is dependent with different crime types.