- The independent variable is the
nature of the parents' relationship. One possible operational
definition would be to record whether or not the parents are divorced; this would make 'parents' relationship' a qualitative
variable. The dependent variable would be attitude towards marriage,
which could be measured on a 10-point scale. This would make
attitude a quantitative variable. The experiment is observational.
|
|
- a) Low arousal: Mean = 5.1; Median = 5; Mode = 5; s2 = 2.32; s = 1.52
Med Arousal: Mean = 17; Median = 17.5; Mode = 18; s2 = 2.0; s = 4.0
High Arousal: Mean = 7.7; Median = 8; Mode = 8; s2 = 1.57; s = 1.25
b)The data suggest that a moderate level of arousal is optimal.
This idea has been borne out by many experiments and is often
referred to as the Yerkes-Dodson law.
- These data suggest that all three distributions are symmetrical (or at least pretty darn close).
|
|
a) Area in tail = .0582
b) Area in tail = .0582
c) Area in body (0.24): .5948
Area in tail (1.98): .0239
.5709
d) Area in tail (0.24): .4052
Area in tail (1.98): -
.0239
.3813
z = x-mean / sd
z = 8 - 12 / 2.5
z = -4 / 2.5
z = -1.6 Area
in tail: .0548
Therefore, you are a relatively industrious individual, as only
about 5.5% of college students procrastinate less than you. |
|
1. Researchers have developed a filament that should add to the life expectancy of light bulbs. The standard 60-Watt bulb burns for an average of mu = 750 hours with sigma = 20. A sample of 100 bulbs is prepared using the new filament. The average life for this sample is 820 hours.
- 80% CI = 820 +/- [1.28 *(20/sqrt(100)] = 820 +/- 2.56 = [817.44 - 822.56]
- 99% CI = 820 +/- [2.58 *(20/sqrt(100)] = 820 +/- 5.16 = [814.84 - 825.16]
2. A survey of students produced the following data regarding their age when they first consumed an alcoholic drink: 11, 13, 14, 12 and 10.
- 95% CI = 12 +/- [2.776 *(1.58/sqrt(5)] = 12 +/- 1.96 = [10.04 - 13.96]
- 99% CI = 12 +/- [4.604 *(1.58/sqrt(5)] = 12 +/- 1.96 = [8.75 - 15.25]
- The Confidence Interval gets wider as the level of confidence increases.
|
|
- Ho: Mu = 1250 Ha: Mu does not = 1250
RR: t < -2.00; t > 2.66
Zobs = x-MUo / [s/(sqrt(n)] = 1284 - 1250 / [90/sqrt(61)] = 2.95
tobs falls in the rejection region. Therefore, we would REJECT
the null. Interpretation: Athletes are drawn from a population
with a higher mean average SAT score than the general student
body.
- Ho: Mu = 1 Ha: Mu does not = 1
RR: t < 2.776; t > 2.776
tobs = x-MUo / [s/(sqrt(n)] = 1.34 -1 / [.5 /sqrt(5)] = 1.52
tobs does not fall in the rejection region. Therefore,
we would FAIL TO REJECT the null. Interpretation: I should
probably stop taking the drug because there is no evidence that nandrolene affects my publication rate (and there might be some negative side effects).
- Ho: Mu = 2.6 Ha: Mu does not = 2.6
RR: t < -2.445; z > 2.445
M = 4
s = 2.08
tobs = M-MUo / [s/(sqrt(n)] = 4 - 2.6 / [2.08/sqrt(7)] = 1.78
tobs does not fall in the rejection region. Therefore, we would FAIL TO REJECT the null hypothesis. Interpretation: I do not remember more or fewer dreams than the average person according to Susan Boyle's book.
|
- Ho: Uhp = Ue Ha: Uhp is not equal to Ue tcrit = 1.98 (use value for df = 120)
SE = sqrt[(502/100) + (752/100)] = 9.01
t = 165-150 / 9.01 = 1.66
tobs is less than tcrit. Therefore, we would fail to reject the
null.
Interpretation: neither printer is faster than the other.
1a) tcrit = 1.658. 1.66 is greater than 1.658 so we would reject the null.
- 95% CI = 15 +/- 1.98 * 9.01 = [-2.84-32.84]. Yes, it makes
sense. We failed to reject the null, which is consistent with
the fact that the interval includes the value 0.
|
|
- tobs = (69.4 - 66.2) / (12.2 /sqrt(120))
- = 2.87 < tcrit (2.66)
-
Therefore, you would reject the null and conclude that the jingle
improved the consumers' moods.
- 99% CI = 3.2 +/- 2.66 * (12.2 /sqrt(120))
- = 3.2 +/- 2.66 * 1.11
- = 3.2 +/- 2.95
- = 3.2 +/- [.25 - 6.15]
-
In other words, we would expect that the average person's mood
would increase by between approximately .25 and 6.15 points after
listening to the jingle.
|
|
-
|
East |
West |
Same |
|
|
2 |
6 |
1 |
|
|
1 |
4 |
0 |
|
|
3 |
6 |
1 |
|
|
3 |
8 |
1 |
|
|
2 |
5 |
0 |
|
|
4 |
7 |
0 |
|
|
|
|
|
Sum |
Sum (x) |
15 |
36 |
3 |
54 |
Sum (x2) |
43 |
226 |
3 |
272 |
(T2)/n |
37.5 |
216 |
1.5 |
255 |
Sum (x2)-(T2)/n |
5.5 |
10 |
1.5 |
17 |
Source |
SS |
df |
MS |
F |
Between |
93 |
2 |
46.5 |
41.03 |
Within |
17 |
15 |
1.13 |
|
Total |
110 |
17 |
|
|
The observed value of F exceeds the critical value. Therefore, we would reject the null and conclude that direction of travel does influence the severity of jet lag.
- Tukey's HSD = q * sqrt(MSE/n)
=
3.67 * sqrt (1.13/6)
= 3.67 * .434 =
1.59 Based
on Tukey's HSD, I would conclude that traveling West is more difficult than either traveling East or staying within the same time zone. Traveling East is also more difficult than staying within the same time zone.
|
|
- In a repeated measures experiment, the same units
(people) serve as observations (subjects) for each level of the
independent variable. In a between subjects experiment, each
level of the independent variable consists of a unique set of
observations.
- The difference between RM- and BS-ANOVA relates to the
error term. In a BS-ANOVA, the error term consists of all of
the variability within each level of the independent variable.
In a RM-ANOVA, the variability within each level of the IV is
derived from two sources: chance variation (error), and differences
between the subjects that comprise the sample. Thus, we can reduce
the error term in our F-ratio by partitioning out the variability
due to individual differences. Reducing the error term in our
F-ratio increases the value of F, which increases the chance
that we will reject the null hypothesis. Because reducing the
error term increases the probability of rejecting the null without
affecting alpha, we say that our test is more powerful.
|
|
Omnibus Test |
Source |
df |
SS |
MS |
F |
p-value |
Model |
8 |
144 |
18.00 |
3.6 |
.0372 |
Error |
99 |
495 |
5.00 |
|
|
Total |
107 |
639 |
|
|
|
Individual Effects |
Source |
df |
SS |
MS |
F |
p-value |
A |
2 |
28.0 |
14.0 |
2.8 |
.0496 |
B |
2 |
16.0 |
8.0 |
1.6 |
.3428 |
AxB |
4 |
100 |
25.0 |
5.0 |
.0215 |
- a) See above.
b) 3 (b/c dfA= 2)
c) 3 (b/c dfB = 2)
d) 108 (b/c dfTotal = 108)
e) 12 (N/treatments = 108 / 9)
f) A is significant; B is not significant; the interaction is
significant.
|
Old AM |
Old PM |
Young AM |
Young PM |
|
Mean |
4 |
2 |
4 |
9 |
|
Sum (x) |
24 |
12 |
24 |
54 |
114 |
Sum(x2) |
110 |
28 |
118 |
498 |
754 |
SS |
14 |
4 |
22 |
12 |
52 |
(G2)/N |
|
|
|
|
541.5 |
(T2)/n |
96 |
24 |
96 |
486 |
702 |
Sum(x2)-(T2)/n |
14 |
4 |
22 |
12 |
52 |
Source |
SS |
df |
MS |
F |
Fcrit |
Model |
160.5 |
3 |
53.5 |
20.58 |
3.10 |
Error |
52 |
20 |
2.6 |
|
|
Total |
212.5 |
23 |
|
|
|
Because the observed value of F exceed the critical value, we conclude that at least one of our treatment means differs from the others (F 3, 20) = 20.58, MSE = 2.6). Now, on to test the main and interaction effects.
|
Older |
Younger |
AM |
PM |
|
Age |
108 |
507 |
|
|
615 |
|
|
|
|
|
|
Test Time |
|
|
192 |
363 |
555 |
Source |
SS |
df |
MS |
F |
Fcrit |
Age |
73.5 |
1 |
73.5 |
28.27 |
4.35 |
Test Time |
13.5 |
1 |
13.5 |
5.19 |
|
Age x Test Time |
73.5 |
1 |
73.5 |
28.27 |
|
The main effect of Age was significant (F (1, 20) = 28.27, MSE = 2.6). Younger adults performed significantly better than older adults. The main effect of Test Time was also significant (F (1, 20) = 5.19, MSE = 2.6). the subjects performed better in the afternoon than in the morning. Finally, the interaction effect was also significant (F (1, 20) = 28.27, MSE = 2.6). Whereas older adults tended to perform better in the morning, younger adults tended to perform better in the afternoon. Moreover, older adults performed just as well as younger adults in the morning, but performed much worse than younger adults when tested in the afternoon.
|
|
- Calculate the regression equation.
SP = S(xi)(yi) - [S(xi)S(yi)] / n
-
= 15,013 - [(109.5*11,087)/77]
-
= 15,013 - 15766.58 = -753.58
SSx = Sxi2 - [(Sxi)2 / n]
-
= 246.25 - [(109.52)/77]
-
= 246.25 - 155.72 = 90.53
B1 = SP / SSx
-
= -753.58 / 90.53 = -8.32
B0 = My - B1 (Mx)
-
= (11,087/77) - (-8.32)*(109.5/77)
-
= 143.98 - (-8.32)*1.42 = 155.82
y = 155.82 - 8.32*x
- The slope tells us the change in y for a unit change in
x. Therefore, if you ate one additional serving of vegetables,
I would expect you to lose 8.32 pounds.
- Use the regression equation with 3 substituted for x.
-
y = 155.82 - 8.32*x
-
y =155.82 - 8.32*(3)
-
y =155.82 - 24.96
-
y =130.86
- Performing the hypothesis test.
-
SSy = Syi2 - [(Syi)2 / n]
-
= 1666795 - (110872/77)
-
= 1666795 - (122921569/77)
-
= 1666795- 1596384.012987
-
= 70410.987013
- SSE = SSy - b1(SP)
-
= 70410.987013 - (-8.32 *-753.58)
-
= 70410.987013 - 6269.7856
-
= 64141.201413
- s2 = MSE = SSE / (n-2)
-
= 64141.201413 / (77-2) = 855.22
s = sqrt(855.22) = 29.24
- t = b1 - 0 / s / sqrt(SSx)
-
= -8.32 / (29.24 / sqrt(90.53))
-
= -8.32 / (29.24 / 9.515)
-
= -8.32 / 3.073
-
= -2.71
Reject the null. We conclude that there is a significant relationship
between weight and vegetable consumption.
- Calculating the correlation coefficient
-
SSy = Syi2 - [(Syi)2 / n]
-
= 1,666,975 - (11,0872/77)
-
= 1,666,795 = 1,596,384.01 = 70,410.99
- r = SP / [sqrt(SSx)(SSy)]
-
= -753.58/(sqrt(90.53*70,410.99))
-
= -753.58 /sqrt(6374306.65) = -.30
r2 = -.302 = .09.
Although vegetable consumption does help us predict weight,
it can only explain 10% of the variance in our dependent measure.
Clearly, other factors are involved
- y = 6.967 + .596(x)
-
.596 hours later.
-
y = 14.119 (approximately 2:00 AM)
-
The tobs = 5.212; because my p-value is less than .05, I would
reject the null and conclude that weekday bedtime is a significant
predictor of weekend bedtime.
-
r = .53; therefore, r2 = .281. This suggests that we can explain
almost 30% of the variance in weekend bedtimes if we know weekday
bedtimes. That is a reasonably high value for "real"
data.
|
|
Solution
x (rt) |
y (errors) |
x2 |
y2 |
s*y |
184 |
10 |
33856 |
100 |
1840 |
213 |
6 |
45369 |
36 |
1278 |
234 |
2 |
54756 |
4 |
468 |
197 |
7 |
38809 |
49 |
1379 |
189 |
13 |
35721 |
169 |
2457 |
221 |
10 |
48841 |
100 |
2210 |
237 |
4 |
26169 |
16 |
948 |
192 |
9 |
36864 |
81 |
1728 |
|
|
|
|
|
1667 |
61 |
350385 |
555 |
12308 |
- SSx = 350385 - (16672/8)
- = 3023.875
-
SSy = 555 - (612/8)
- = 89.875
- SP = 12308 - (1667*61/8)
- = -402.875
- r = SP / sqrt(SSx*SSy)
- = -402.875 / sqrt(3023.875*89.875)
- = -.7728
- Thus, there is a strong negative relationship between the
speed of response and accuracy. In other words, faster responses
tend to produce more errors.
- 2) Yes. The correlation coefficient is significant; r (6) = -.773, p = .025).
|
|
- Yes, the overall model will help SMS predict success
in the fraternity. This conclusion is based on the fact that
the omnibus ANOVA leads us to rejct the null (F (6, 42_ = 6.00,
MSE = 5.00, p = .0125).
- R-squared = 180/360 = .50
- y = 22 - .16(GPA) + 2.17(Math-GPA) + 4.81(Protect) + 6.03(Trek)
+ .88(T-shirts) - 3.25(Friends)
- y = 22 - .16(3.20) + 2.17(3.80) + 4.81(2) + 6.03(2) + .88(12)
- 3.25(2)
- y = 22 - .512 + 8.246 + 9.62 + 12.06 + 10.56 - 6.5
y = 55.474
- Therefore, we would expect your brother to be admitted
to the fraternity.
|
|
|
|
|