5-6 Page Applied Decision Method Paper...Effecs on Gas mileage a Staistical Model. Please be familiar with regression models
pepperkitty9
notes_project_revised.docxI ran 18 vehicles all of them cars in increments of five years from 1975 to 2015 with front wheel drive and rear wheel drive in a new model.
In the first single regression model, I ran the independent variable of model year to witness any relationship on gas mileage. The multiple r factor suggests a moderate correlation (59%) between model year and MPG. The model is valid as is proven by the extremely low p factor and significance f as they are well below 5%. The r squared is lower than we’d like to see if we were to make the statement that there is indeed a definitive relationship between the two variables, suggesting that the variability in MPG are only explained by model year about 35 % of the time. The Year coefficient suggests that for every increase in year (I year at a time) there is an increase in MPG by 0.17.
The next model I ran was weight, which once again surprises me with a model that is invalid. The high significance f and p factor are above 5% suggesting a null hypothesis. Plus, the correlation between the two factors are extremely low (multiple r and r square)
The model run with FWD and RWD were run using dummy variables and is a valid model with the low significance f and p factor. The multiple r shows a 68% correlation between FWD/RWD and MPG however once again the r square was lower explaining only 46% of the variability in MPG due to FWD or RWD. The X coefficient shows that FWD vehicles get about 5.1 MPG better than RWD vehicles. And if I’m reading the Intercept Coefficient correctly, the MPG would average 19.1 mpg for RWD vehicles. Since this is a model using dummy variables, the intercept coefficient only outlines the mean for the reference group (RWD) since it is the y intercept when x=0 hence when x=0 were talking about RWD, when x=1 we are talking about FWD. Anyway, I digress. Let’s continue….
I then ran a regression using horsepower as the independent variable which shocked me as it shows no correlation whatsoever between the variables and the hypothesis is completely null with a ridiculously high significance f and p factor as well as terribly low r square and multiple r factors.
The multiple regression that I finally ran shows that it is a sound hypothesis and a statistically valid model, However the p values of the variables changes significantly in this model, here the FWD/RWD becomes extremely high showing that it is not a valid variable in the model, while year, weight, and horsepower all have significantly lower p values highlighting significance in this model. The significance f is also extremely low highlighting the validity of the model and hypothesis. The multiple r shows a 97% correlation of the variables and the MPG and the adjusted r squared shows a 93% explanation in the variability of MPG due to the variables. In looking at the X coefficients there is an inverse relationship with the weight, horsepower, and FWD variables, showing that as they increase the MPG seems to slightly decrease (.005mpg for weight and .035 for horsepower) while the model year has a positive relationship increasing MPG as years go by (.374). I am at somewhat of a loss as to why the FWD was so significant in previous single models and why now it is null in the multiple model. Furthermore, the changes in the other previously insignificant variables are now notable…..
