Estimated Prius Gas Mileage

Methods

Each time I filled up the gas tank, I collected the number of gallons that were dispensed until the first automatic shutoff. I wrote down the estimated mpg, the trip odometer reading since the last fill-up, and the overall odometer reading. I calculated the mpg and mpg difference from the estimated mpg. I used a paired t test to see if there was a difference between the estimated and actual mpg.

library(dplyr)

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

x <- read.csv("../datasets/mpg.csv")
x <- tbl_df(x)
x

## # A tibble: 5 x 6
##         date miles_total miles_trip mpg_est price gallons
##       <fctr>       <int>      <dbl>   <dbl> <dbl>   <dbl>
## 1 2017-07-20       90208      358.1    48.7 19.54   6.860
## 2 2017-08-19       90596      387.5    53.7 26.02   8.677
## 3 2017-09-03       90984      383.5    55.9 22.10   7.730
## 4 2017-09-28       91748      359.3    51.6 21.82   7.796
## 5 2017-10-13       92216      467.0    54.6 26.08   9.029

Results

x <- x %>% mutate(mpg_calc = miles_trip/gallons) %>% mutate(mpg_diff = mpg_calc - mpg_est)

x %>% select(date, mpg_est, mpg_calc, mpg_diff)

## # A tibble: 5 x 4
##         date mpg_est mpg_calc  mpg_diff
##       <fctr>   <dbl>    <dbl>     <dbl>
## 1 2017-07-20    48.7 52.20117  3.501166
## 2 2017-08-19    53.7 44.65829 -9.041708
## 3 2017-09-03    55.9 49.61190 -6.288098
## 4 2017-09-28    51.6 46.08774 -5.512263
## 5 2017-10-13    54.6 51.72223 -2.877772

From looking at the data, the estimated mpg from the car’s dashboard was higher than the calculated mpg based on the odometer and gallons readout on four out of five occasions. It’s not clear why that first one was different, but I included it since I didn’t have a good reason to throw it out.

The results of the paired t-test are as follows.

t.test(x$mpg_diff)

## 
##  One Sample t-test
## 
## data:  x$mpg_diff
## t = -1.9014, df = 4, p-value = 0.13
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
##  -9.948447  1.860977
## sample estimates:
## mean of x 
## -4.043735

Discussion

There was not a significant difference between the estimated and calculated mpgs. If the first data point is removed, then there is a significant difference. Because other sources on the internet suggest that there is a quite significant difference, I’ll keep collecting more data and update the analysis in a few months.