Apr 01, 2015
Descriptiono We chose to do our project on cars in
South's parking lotto see if our parking lot could be considered representative of the entire population of cars
o We studied many aspects of them like color, size, type, etc. to try and make conclusions about the population
o We went outside in the parking lot to explore!
History of Cars
o 1769- First self-propelled caro 1886- Internal combustion engines
developedo 1896- First road traffic deatho 1997- Green cars
Taking the Sample
o Every 5 cars in CB South parking lot
o More variables to decide from Color Type (Car, SUV, Truck, Other) Number of Doors
Sampling Results
Red SilverGreen Blue White Black Other Total
2-Door Car 3 3 1 0 4 2 1 144-Door Car 6 20 7 5 7 7 5 57
SUV 2 6 3 4 5 3 1 242-Door Truck 0 1 0 1 0 1 1 44-Door Truck 0 0 0 0 1 2 0 32-Door Other 1 0 0 0 0 0 1 24-Door Other 3 2 1 3 2 0 1 12
Total 15 32 12 13 19 15 10 116
Cars SUVs Trucks Other0
10
20
30
40
50
60
70
80
Type Of Vehicle
Type Of Vehicle
Nu
mb
er
Of
Veh
icle
s
Cars
SUVs
Truc
ks
Oth
er0
10
20
30
40
50
60
70
80
Type Of Ve-hicle
Type Of Vehicle
Nu
mb
er
Of
Veh
icle
soChose to this graph to
display the overall results in a simple and general form
oMajority of vehicles are cars
oFew trucks, but may be different in population where more workers use trucks to transport heavy items
0
5
10
15
20
25
30
35Total Color of Vehicles
OtherTrucksSUVsCars
Color Vehicle
Nu
mb
er
Of
Veh
icle
s
0
5
10
15
20
25
30
35 Other
Trucks
SUVs
Cars
Color Vehicle
Nu
mb
er
Of
Veh
icle
s
Colorso Used this graph because it’s more specific but not too specifico Shows silver is the main color of the populationo Most colors have about equal amounts of SUVs
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%Total Color of Vehicles
Other
Trucks
SUVs
Cars Color Vehicle
Nu
mb
er
Of
Veh
icle
s
Red Silver Green Blue White Black Other0
5
10
15
20
25
30
35
Color and Type4-Door Other2-Door Other4-Door Truck2-Door TruckSUV
Color Vehicle
Nu
mb
er
of
Ve-
hic
les
Red SilverGreen Blue White Black Other0
5
10
15
20
25
30
35
4-Door Other2-Door Other4-Door Truck2-Door TruckSUV4-Door Car2-Door Car
ColorN
um
be
r
Color and Type
o Stacked bar grapho Shows all variableso Silver 4- door cars are
dominanto Blue has no 2- door
cars, so in the population they must be minimal
Tests We Usedo Chi- Square Test: Goodness of
Fito Uniformo Our Sample vs. North America
o Chi- Square Test for Association: Color vs. Car Size
o One Proportion Z Test for SUVso One Proportion Z Interval for
SUVs
Assumptions for Goodness of Fit
o Assumed (we performed the test)
o √ (all expected counts are 16.5714)
o SRSo Sample size large enough that all expected counts are greater than or equal to 5
2
2 Goodness Of Fit
We found a website releasing the car colors of 2006 in North America from the DuPont Annual Color Popularity Report. We decided to see if this distribution (“overall”) fit the distribution of car colors from CB South’s parking lot. We included light brown and yellow/gold in the “other” category, making “other” 11%. We included silver and gray together, as we had in our study, making “silver” 32%. We also combined white pearl with white, making “white” 19%.
Test – Goodness of Fit
Expected: Red - (.11)(116) = 12.76Silver - (.32)(116) = 37.12Green - (.04)(116) = 4.64Blue - (.11)(116) = 12.76White – (.19)(116) = 22.04Black – (.13)(116) = 15.08Other – (.11)(116) = 12.76
Observed: Red - 15Silver - 32 Green - 12Blue - 13White - 19Black - 15Other - 10
Ho: The observed frequency distribution of car colors fits the expected.Ha: The observed frequency distribution of car colors does not fit the expected.
Check• Assumed• Does not check- Since green’s expected
count of 4.64 is close to 5 we proceed
State• SRS• Sample size large enough so all
expected counts ≥ 5
12.37
)12.3732(
76.12
)76.1215(
exp
)exp( 2222
ected
ectedobserved …= 13.7952
2
0320.)7952.13( 2 p
df=6α=.05
Test – Goodness of Fit2
We reject Ho in favor of Ha because p-value of .0320 < α =.05. We have sufficient evidence that the observed frequency distribution of
car colors does not fit the expected distribution.
The category of green seemed to be the category furthest off. We suspected that if green were not part of the test, we might have failed to reject. Even though our p-value was still less than .05, it was much higher
than our p-value in the previous test, which was .0032.
We could have improved this test by having a sample size large enough that all expected counts were greater than 5. Perhaps we could have included green in the “other” group in order to avoid this problem, or we could have increased
our sample size (every fourth car instead of every fifth, for example).
0320.)7952.13( 2 p
Ho: The observed frequency distribution of car colors fits the expected.Ha: The observed frequency distribution of car colors does not fit the expected.
2 Test of AssociationState• Two Independent SRS• Sample size large enough that all
expected counts ≥ 5
Check• Assumed• Does not check- we choose to proceed
Ho: Car color and size are independent.Ha: Car color and size are dependent.
Red Silver
Green
Blue White
Black Other
Small 9.181 19.586
7.345 7.957 11.629
9.181 6.1207
Large 5.819 12.414
4.655 5.043 7.371 5.819 3.879
586.19
)586.1923(
181.9
)181.99(
exp
)exp( 2222
ected
ectedobserved
5922.)629.4( 2 p
…=4.629
df=6 α=.05
2 Test of Association
We fail to reject Ho in favor of Ha because p-value of .5922>α=.05. We have sufficient evidence that car color and size are independent. o Chose to this test to see whether or not there was an
association between the size of the car and its coloro Thought that it was more likely for a small, sporty car to be a
flashy color like red rather than a large truckoGrouped our sample into two categories: oSmall, for cars, and large, for SUVs, trucks, and otheroBecause the samples of SUVs, trucks, and other vehicles were
too small to be statistically viable on their own. ***We concluded that there was no association between car color and size. Thus, it is no more or less likely for a large car to be a certain color than a small car. ***
Red
Silv
er
Green
Blue
Whi
te
Black
Other
05
10152025
Size vs. Color
Small
Large
Color
Nu
mb
er
of
Cars
Comparing Size and Color
o Displays the amount of cars of each color
o Stacked bar graph helps us easily compare the car sizes to the color.
For the most part, the number of small cars of each color is greater than the number of large cars because our sample of small cars was so much greater. The exception is blue, where the number of large cars of that color is greater than the number of small cars.
One Proportion Z Test for SUVs
Ho: p=.3Ha: p<.3
State• SRS• np ≥10• n(1-p)
≥10• pop ≥
10n
Check• Assumed• 116×.3 ≥10• 116×.7 ≥10• pop ≥1,160
1882.2)1(
ˆ
npp
ppz
01433.)1882.2( zPWe reject Ho in favor of Ha because the p-value of .01433 is greater than α=.05. We have sufficient evidence that the proportion of SUVs is less than .30.
We decided to do this test because we felt that the proportion of SUVs was almost .5, so we used .3 because we knew the proportion wouldn’t be that close to .5. We felt that the proportion of SUVs would be less than .3 because they use a lot of gasoline and the majority of people are trying to become more eco-friendly.
One Proportion Z Interval for SUVs
Confidence level of 96%
State• SRS• np ≥10• n(1-p)
≥10• pop ≥
10n
Check• Assumed• 116×.3 ≥10• 116×.7 ≥10• pop ≥1,160
)19329,.06533(.)ˆ1(ˆ
*ˆ
n
ppzp
1293.ˆ p
We are 96% confident that the proportion of SUVs lies between .06533 and .19329.
We decided to do interval because we wanted to find out where the actual proportion of SUVs fell. By doing a confidence interval we found that the proportion of SUVs is not that high compared to what we originally thought. People really are becoming more eco-friendly!
Mistakes
o Not every car in the parking lot is there everyday, so we probably missed someo We did not meet all the assumptions, but we continued with the testsoWe should have picked the cars by using a software to randomize which cars we looked at by their spot number
Our Conclusionso Expected the main colors to be black and whiteo Large amount of red was surprisingo Silver is very popularo School parking lot doesn’t seem to match national
distribution of carso Could have improved chi- square test by increasing the
sample size so that all expected counts were greater than or equal to five.o Every 4 cars instead of 5o Stratified random sampling would have worked better
in this caseo Split the cars in the parking lot into small cars and
large cars and taken a simple random sample of the small cars and a simple random sample of large cars. Therefore, both samples would be sufficiently large enough for the assumptions to pass.
Our Opinions
o Too much work in too little time!o Way too cold and should have found a
different way to sampleo Liked looking at cool car pictureso Glad it’s over