Вы находитесь на странице: 1из 8

Justin Zweig

Dr. Raman
ISE 4553
112979659
Assignment #4
1.) Do iPhone generation has the best battery life durability? I would set up this
experiment by collecting 100 replicates from 4 different factor levels. The factor levels
would be the last 4 generations of iPhone: 7, 6s, 6, and the 5s.
The 2 questions I would ask iPhone owners would be: How long have you had your
iPhone and what is the average battery life after a full battery charge? From these 2
questions I would plot a scatter plot with a regression line of each different iPhone
measuring time and the average life expectancy of the battery. From the graphs, I would
hope to see if iPhone batteries have got worse or better over the generations. A factor that
could affect the quantitative data could be whether the iPhone owner is running their
phone in battery saver mode or how many apps the iPhone is running in idle mode.
Ho: u7=u6s=u6=u5s
H1: u7u6su6u5s+
From here, I would average the battery lives of each iPhone generation that have been
used for over a year. I chose 1 year because I personally think that 1 year gives a phone
battery enough time to wear down. I realize there may not be enough time to carefully
measure the iPhone 7 because it just released, I would probably delay the experiment or
not include iPhone 7 data if I were to conduct the experiment today.
2.)
a.) A measurable dependent variable in this experiment is the average maintenance cost
of each 737 over the course of a year.
b.) A factor that could affect the independent variable could be the age of the plane itself
or the amount of trips the plane has been on. Three levels that could be measured:
plane engine maintenance, body maintenance, and other maintenance.
3.)
3.1)

V1 = (a-1) = (4-1) = 3
V2 = (N-a) = ((4*6) - 4) = 20
For alpha = 0.05, f(0.05, 3, 20) = 3.0984
For alpha = 0.025 f(0.025, 3, 20) = 3.8587
0.025 < actual p-value < 0.05
> 1- pf(3.26, 3, 20)
[1] 0.04300116

3.2)

V1 = (a-1) = (6-1) = 5

V2 = (N-a) = ((6*3) 6) = 12
For alpha = 0.01, f(0.01, 5, 12) = 5.064
For alpha = 0.005, f(0.005, 5, 12) = 6.07
0.005 < actual p value < 0.01
> 1- pf(5.81,5,12)
[1] 0.005941688

3.3)
Source
Factor
Error
Total

Df
3
16
19

SS
36.15
159.89
196.04

MS
12.05
9.993

F
1.205

P
0.334

SS
987.71
186.53
1174.24

MS
246.93
7.46

F
33.1

P
1.18e-09

> 1 - pf(1.20583, 3, 16)


[1] 0.3395229

3.4)
Source
Factor
Error
Total

Df
4
25
29

1 - pf(33.1, 4, 25)
[1] 1.182879e-09

4.)
Ho: u1=u2=u3
H1: u1u2u3

Arrangement has no effect on reaction time


Arrangement has an effect on reaction time

> InstrumentsData <- read.csv("~/ISE 4553 data/InstrumentsData.csv")


> View(InstrumentsData)
> InstrumentsData$Arrangement <- as.factor(InstrumentsData$Arrangement)\
entsData)
>myANOVA <- aov(Reaction.time~Arrangement, data =InstrumentsData)
> summary(myANOVA)
Df Sum Sq Mean Sq F value Pr(>F)
Arrangement 2 81.43 40.71 11.31 0.000318 ***
Residuals 25 90.00 3.60
--Signif. codes:
0 *** 0.001 ** 0.01 * 0.05 . 0.1 1

The p-value is less than alpha=0.01, the level of significance. Reject the null hypothesis, the
arrangements have an effect on reaction time.
> print(model.tables(myANOVA,"means"))
Tables of means
Grand mean
10.14286
Arrangement

arrangement1 arrangement2 arrangement3


12.5
10
8
rep
8.0
12
8

Based off of the Normal Q-Q plot, we can assume errors are normally and independently
distributed with a mean zero. Nothing unusual looking from the residuals vs fitted, we can
assume constant variance.
>TukeyHSD(myANOVA, conf.level = 0.99)
Tukey multiple comparisons of means
99% family-wise confidence level
Fit: aov(formula = Reaction.time ~ Arrangement, data = InstrumentsData)
$Arrangement
diff
lwr
arrangement2-arrangement1 -2.5 -5.272334
arrangement3-arrangement1 -4.5 -7.536939
arrangement3-arrangement2 -2.0 -4.772334
upr
p adj
arrangement2-arrangement1 0.2723335 0.0208576
arrangement3-arrangement1 -1.4630608 0.0002074
arrangement3-arrangement2 0.7723335 0.0728959

Means of arrangements 2-1 and arrangements 3-2 are statistically the same because its p-value is
greater than the significance level. Arrangements 3-1 are statistically different. The arrangements
is best to worst order is: Arrangement 3, Arrangement 2, and Arrangement 1.
5.)
Ho: u1=u2=u3=u4
H1: u1u2u3u4
>
>
>
>

the mixing techniques are the same


mixing technique has an effect on tensile strength

problem5 <- read.csv("~/ISE 4553 data/problem5.csv")


View(problem5)
problem5$Mixing.Techniques <- as.factor(problem5$Mixing.Techniques)
anova5 <- aov(Tensile.Strength~Mixing.Techniques, data=problem5)

> summary(anova5)
Df Sum Sq Mean Sq F value
Mixing.Techniques 3 489740 163247 12.73
Residuals
12 153908 12826
Pr(>F)
Mixing.Techniques 0.000489 ***
Residuals
--Signif. codes:
0 *** 0.001 ** 0.01 * 0.05 . 0.1 1
0.226 0.877
Residuals
12 609268 50772

P-value is less than the significance level of 0.05, reject the null. Technique has an effect on
tensile strength.
> print(model.tables(anova5,"means"))
Tables of means
Grand mean
2931.812
Mixing.Techniques
technique1 technique2 technique3 technique4
2971.0
3156.2
2933.8
2666.2

The spread of data in the residuals vs fitted plot is roughly normal looking and can assume
constant variance. Errors in the normal Q-Q plot seem normally distributed with a mean 0 and
can assume normality.
> TukeyHSD(anova5)
Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = Tensile.Strength ~ Mixing.Techniques, data = problem5)

$Mixing.Techniques
diff
technique2-technique1
technique3-technique1
technique4-technique1
technique3-technique2
technique4-technique2
technique4-technique3
upr
technique2-technique1
technique3-technique1
technique4-technique1
technique3-technique2
technique4-technique2
technique4-technique3

lwr
185.25 -52.50029
-37.25 -275.00029
-304.75 -542.50029
-222.50 -460.25029
-490.00 -727.75029
-267.50 -505.25029
p adj
423.00029 0.1493561
200.50029 0.9652776
-66.99971 0.0115923
15.25029 0.0693027
-252.24971 0.0002622
-29.74971 0.0261838

6.)

The interval includes 0, answer stays in the same in problem 5. There is not a statistical
difference between mean treatments in 1 and 3.
7.)
Ho: uH=uE=uT
H1: uHuEuT
healthcare

Each fund is essentially the same in terms of YTD return


The funds are different in YTD return
energy

technology

-3.87

15.87

-19.97

9.94

-0.5

21.88

-0.86

18.03

-2.14
15.0
2
10.4
2
7.10
5

Anova: Single Factor


SUMMARY
Groups

Count

Column 1

Column 2

Column 3

ANOVA
Source of
Variation
Between
Groups
Within
Groups

SS
1050.4
59
488.82
5

Total

1539.2
84

Sum
-25.2
65.7
2
30.4
05

df
2
9

Averag
e
-6.3
16.43
7.6012
5

MS
525.22
97
54.313
89

Varian
ce
85.335
8
24.898
73
52.707
14

F
9.6702
64

Pvalue
0.0057
31

F crit
4.2564
95

11

The p-value of 0.006 is less than the level of significance, therefore, reject the null hypothesis
and means of each category is statistically different.

> TukeyHSD(anova7)
Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = YTD ~ Category, data = problem7)
$Category
diff
lwr
healthcare-energy
22.73000 8.1802093
technology-energy
13.90125 -0.6485407
technology-healthcare -8.82875 -23.3785407
upr
p adj
healthcare-energy
37.279791 0.0046436
technology-energy
28.451041 0.0607009
technology-healthcare 5.721041 0.2592942

Healthcare and energy categories are statistically different. I would rank these funds from best to
worst: Energy, technology, and healthcare.
8.)

Вам также может понравиться