Question no 1.

Descriptives
Std.
type Statistic Error
strengt 1 Mean 49.5333 1.80440
h 95% Confidence Lower 45.6633
Interval for Mean Bound
Upper 53.4034
Bound
5% Trimmed Mean 49.5370
Median 50.0000
Variance 48.838
Std. Deviation 6.98843
Minimum 37.00
Maximum 62.00
Range 25.00
Interquartile Range 10.00
Skewness -.060 .580
Kurtosis -.530 1.121
2 Mean 74.9375 2.27022
95% Confidence Lower 70.0986
Interval for Mean Bound
Upper 79.7764
Bound
5% Trimmed Mean 75.2083
Median 76.5000
Variance 82.463
Std. Deviation 9.08089
Minimum 55.00
Maximum 90.00
Range 35.00
Interquartile Range 10.75
Skewness -.342 .564
Kurtosis .376 1.091
3 Mean 52.5385 2.59836
 95% Confidence Lower 46.8771
Interval for Mean Bound
Upper 58.1998
Bound
5% Trimmed Mean 52.1538
Median 51.0000
Variance 87.769
Std. Deviation 9.36852
Minimum 39.00
Maximum 73.00
Range 34.00
Interquartile Range 13.00
Skewness .829 .616
Kurtosis .496 1.191
4 Mean 65.3750 2.86484
95% Confidence Lower 59.2687
Interval for Mean Bound
Upper 71.4813
Bound
5% Trimmed Mean 65.1389
Median 65.0000
Variance 131.317
Std. Deviation 11.4593
5
Minimum 47.00
Maximum 88.00
Range 41.00
Interquartile Range 18.25
Skewness .301 .564
Kurtosis -.278 1.091
5 Mean 47.0000 1.69464
95% Confidence Lower 43.3077
Interval for Mean Bound
Upper 50.6923
Bound
5% Trimmed Mean 46.8889
Median 48.0000
Variance 37.333
 Std. Deviation 6.11010
Minimum 36.00
Maximum 60.00
Range 24.00
Interquartile Range 7.50
Skewness .181 .616
Kurtosis .920 1.191
6 Mean 69.9333 2.99757
95% Confidence Lower 63.5042
Interval for Mean Bound
Upper 76.3625
Bound
5% Trimmed Mean 69.5926
Median 68.0000
Variance 134.781
Std. Deviation 11.6095
2
Minimum 54.00
Maximum 92.00
Range 38.00
Interquartile Range 18.00
Skewness .235 .580
Kurtosis -.837 1.121

The mean for type5 is 47 which is the smallest mean as compared to the rest. The mean for type1
distribution is 49.5333, type3 is 52.5385, type4 is 63.3750, and type6 is 69.9333, which are in ascending
order. The distribution with the largest mean is type2 with a mean of 74.9375.

As per standard deviation, the distribution with the smallest deviation is type5 with the value of
6.11010. Type6 has the largest standard deviation of 11.60952 which is nearly double as compared to
the smallest one. Type1 has another smaller standard deviation of 6.98843, and type4 has a standard
deviation of 11.45935 which is similar to the largest standard deviation. In ascending order, type2 has a
standard deviation of 9.08089 and type3 has a value of 9.36852.
B

From the boxplot, we can see that type1 is approximately symmetric. Type2 and type5 are skewed to
the left whereas type3, type4 and type6 are right-skewed.

Type2 has the largest median and type5 has the smallest one. Type1 and type3 have an approximately
similar median and so do type4 and type6.

As per the spread, type5 has the smaller IQR, and type4 and type6 are the two distributions with the
largest IQR. The spreads for type1 and type2 are very similar as well whereas type3 is slightly larger.

There is one outlier (No. 72) in the distribution type5.

A.
ANOVA
strength
Sum of Mean
Squares df Square F Sig.
Between 10046.858 5 2009.372 22.637 <.001
Groups
Within 7278.585 82 88.763
Groups
Total 17325.443 87

Ho: Reduced model: 1 =2 = 3 =4 =5 =6 (All means are equal: One mean model)

Ha: Full model: There are differences. (Not all means are equal: 6 means model)

sums of squared residuals from the full model = SS(f) = 7278.585

sums of squared residuals from the reduced model = SS(r) = 17325.443

The pooled estimate of the variance is 17325.443

F-stat = 22.637

P-value = <0.001

In conclusion, at a significance level of 0.01, with a p-value <0.001 we have extremely strong evidence to
conclude that some alloys are stronger than others.

SS(f) = 7278.585 ; df(f) = 88 – 6 = 82; for f = 6

SS(r) = 17325.443: df(r)= 88 – 1 = 87; for r = 1

F* = [SS(r) - SS(f)] / [df(r) - df(f)] = [SS(1) - SS(6)] / [df(1) - df(6)]

SS(f) / df(f) SS(6) / df(6)

= (17325.443 - 7278.585) / (87 – 82)

7278.585 / 82
= 2009.3716 / 88.76323171
= 22.63743175 ~ 22.637
p value = P (F(5,82)  22.637 ) = <.001
Question No 3.
Which alloys tend to differ in their tensile strength from the others? Answer the question by carrying out
the Tukey’s and Scheffe’s multiple-comparison procedures at the level of significance 0.05. Summarize
significant differences using a means diagram. Compare the conclusions based on the two procedures.

Multiple Comparisons
Dependent Variable: strength
95% Confidence
Mean Interval
(I) (J) Difference Std. Lower Upper
type type (I-J) Error Sig. Bound Bound
Tukey 1 2 -25.40417* 3.38604 <.001 -35.2853 -15.5230
HSD 3 -3.00513 3.57008 .959 -13.4233 7.4131
4 -15.84167* 3.38604 <.001 -25.7228 -5.9605
5 2.53333 3.57008 .980 -7.8849 12.9515
6 -20.40000* 3.44022 <.001 -30.4392 -10.3608
2 1 25.40417* 3.38604 <.001 15.5230 35.2853
3 22.39904* 3.51790 <.001 12.1331 32.6650
4 9.56250 3.33098 .057 -.1579 19.2829
5 27.93750* 3.51790 <.001 17.6716 38.2034
6 5.00417 3.38604 .679 -4.8770 14.8853
3 1 3.00513 3.57008 .959 -7.4131 13.4233
2 -22.39904* 3.51790 <.001 -32.6650 -12.1331
4 -12.83654* 3.51790 .006 -23.1025 -2.5706
5 5.53846 3.69539 .666 -5.2454 16.3223
6 -17.39487* 3.57008 <.001 -27.8131 -6.9767
4 1 15.84167* 3.38604 <.001 5.9605 25.7228
2 -9.56250 3.33098 .057 -19.2829 .1579
3 12.83654* 3.51790 .006 2.5706 23.1025
5 18.37500* 3.51790 <.001 8.1091 28.6409
6 -4.55833 3.38604 .758 -14.4395 5.3228
5 1 -2.53333 3.57008 .980 -12.9515 7.8849
2 -27.93750* 3.51790 <.001 -38.2034 -17.6716
3 -5.53846 3.69539 .666 -16.3223 5.2454
4 -18.37500* 3.51790 <.001 -28.6409 -8.1091
 6 -22.93333* 3.57008 <.001 -33.3515 -12.5151
*
6 1 20.40000 3.44022 <.001 10.3608 30.4392
2 -5.00417 3.38604 .679 -14.8853 4.8770
*
3 17.39487 3.57008 <.001 6.9767 27.8131
4 4.55833 3.38604 .758 -5.3228 14.4395
*
5 22.93333 3.57008 <.001 12.5151 33.3515
*
Scheffe 1 2 -25.40417 3.38604 <.001 -36.9511 -13.8572
3 -3.00513 3.57008 .982 -15.1797 9.1695
*
4 -15.84167 3.38604 .001 -27.3886 -4.2947
5 2.53333 3.57008 .992 -9.6413 14.7079
*
6 -20.40000 3.44022 <.001 -32.1317 -8.6683
*
2 1 25.40417 3.38604 <.001 13.8572 36.9511
*
3 22.39904 3.51790 <.001 10.4024 34.3957
4 9.56250 3.33098 .157 -1.7967 20.9217
*
5 27.93750 3.51790 <.001 15.9408 39.9342
6 5.00417 3.38604 .822 -6.5428 16.5511
3 1 3.00513 3.57008 .982 -9.1695 15.1797
*
2 -22.39904 3.51790 <.001 -34.3957 -10.4024
*
4 -12.83654 3.51790 .028 -24.8332 -.8399
5 5.53846 3.69539 .813 -7.0634 18.1404
*
6 -17.39487 3.57008 <.001 -29.5695 -5.2203
*
4 1 15.84167 3.38604 .001 4.2947 27.3886
2 -9.56250 3.33098 .157 -20.9217 1.7967
*
3 12.83654 3.51790 .028 .8399 24.8332
*
5 18.37500 3.51790 <.001 6.3783 30.3717
6 -4.55833 3.38604 .873 -16.1053 6.9886
5 1 -2.53333 3.57008 .992 -14.7079 9.6413
*
2 -27.93750 3.51790 <.001 -39.9342 -15.9408
3 -5.53846 3.69539 .813 -18.1404 7.0634
*
4 -18.37500 3.51790 <.001 -30.3717 -6.3783
*
6 -22.93333 3.57008 <.001 -35.1079 -10.7587
*
6 1 20.40000 3.44022 <.001 8.6683 32.1317
2 -5.00417 3.38604 .822 -16.5511 6.5428
*
3 17.39487 3.57008 <.001 5.2203 29.5695
4 4.55833 3.38604 .873 -6.9886 16.1053
*
5 22.93333 3.57008 <.001 10.7587 35.1079
*. The mean difference is significant at the 0.05 level.
From the Tukey test, we observe that type1 is different in strength when compared to
type2, type4, and type6 as all their p-value is less than 0.001 which is significantly lower
as compared to our alpha = 0.05.
Similarly, type2 is different from type1, type3, and type5.
Moreover, Type3 is different as compared to type2, type4 and type6, and type4 is
different than type5, type3 and type1 as their p-values are significantly smaller.
Type5 is different than type2, type4 and type6.
Likewise, type6 is different from type1, type3 and type5.

Means Diagram:

Type5 Type1 Type3 Type4 Type6 Type2

From the Scheffe test, when compared at the significance level of 0.05, we can observe
that type1 is different in strength when compared to type2, type4, and type6 as all their
p-value is less than 0.001 which is significantly lower as compared to our alpha = 0.05.
Similarly, type2 is different from type1, type3, and type5. Type2 is similar to type2 and
type6.
Moreover, Type3 is different as compared to type2, type4 and type6.
Type4 is different than type5, type3 and type1 as their p-values are significantly smaller.
Likewise, type5 is different type2, type4 and type6.
Type6 is different than type1, type3, and type5.

Means Diagram:

Type5 Type1 Type3 Type4 Type6 Type2

The end results from both the tests are the same in determining the difference in the alloy’s
comparative strength, the only noticeable difference between the test is p-values for some of the
differences. For example, from the Tukey test the p-value when comparing type2 and type4 is 0 .057
whereas, from Scheff’s test, the p-value for the exact comparison is 0.157. Even after
the difference, the null hypothesis still gets rejected in both cases as the p-value is
larger than our significance level.

Question no 4.
a.
Linear combination of means is:
(1 + 2 + 3 ) – (4 +5 ) = 1/3 + 2/3+ 3/3 - 4/2 - 5/2
3 2

At alpha = 0.02, with a p-value of 0.217, we have very weak evidence to conclude that nickel-based
alloys tend to differ from iron-based alloys in their tensile strength.

b.
Linear combination of means is:
(1 + 2 + 3 ) – (6 ) = 1/3 + 2/3+ 3/3 - 6/1
3 1

At alpha = 0.02, with a p-value of <0.001, we have very strong evidence to conclude that nickel-based
alloys tend to differ from cobalt-based alloys in their tensile strength.

c.
Linear combination of means is:
(4 +5 ) – (6 ) = 4/2 + 5/2 - 6/1
2 1
At alpha = 0.02, with a p-value of <0.001, we have very strong evidence to conclude that iron-based
alloys tend to differ from the cobalt-based alloy in their tensile strength.

Question no 5.
Calculate the 98% confidence intervals for the contrasts defined in question 4. Clearly state the value of
the contrast and its standard error from the output. Find the critical value using the t-table. Then
calculate the intervals by hand. Do the results agree with your test conclusions in question
For 98% confidence interval; alpha/2 = 0.01
a.
Value of contrast ()= 16.8936 / 6 = 2.8156
SE( ) = 13.58505 / 6 = 2.264175
df = n - k = 82
Critical value (t _ 82, 0.01 ) ~ (t_80, 0.01) = 2.378
Confidence interval = ( - t* x SE( ),  + t* x SE( ))
 ( 2.8156 - 2.378 * 2.264175 , 2.8156 + 2.378 * 2.264175)
 (-2.56860815 ,