Solved Exercises Using Statgraphics. Book by Montgomery D.

Three to thirteen) Four chemists are asked to determine the percentage of methanol in
certain chemical compound. Each chemist makes three determinations, and the results are
the following:

Chemicals Percentage of methanol

1 84,99 84.04 84.38
2 85.15 85.13 84,88
3 84.72 84.48 85.16
4 84.20 84.10 84.55

RESPONSE
Statistical Summary for % methanol content
Quimico Recuent Promedi Desviación Coefficient of Mínimo Máximo Rang
s o o Standard Variation o
1 3 84.47 0.481352 0.56985% 84.04 84.99 0.95
2 3 85,0533 0,150444 0.176882% 84,88 85,15 0,27
3 3 84,7867 0,344867 0.406747% 84,48 85,16 0,68
4 3 84,2833 0,236291 0.280353% 84.1 84.55 0.45
Total 12 84.6483 0.415907 0.491335% 84.04 85.16 1.12

Chemical Bias Kurtosis

s Standardized Standardized
1 0.574147
2 -1,20044
3 0.592126
4 0.982621
Total -0.172641 -1.07918

The StatAdvisor
This table shows different statistics of % methanol for each of the 4
levels of Chemicals. The main purpose of the one-way analysis of variance is to
compare the means of the different levels, listed here under the column of
Average. Graphic Selections of Averages from the Graphic Options list to display
graphically the averages.

WARNING: There is more than a 3 to 1 difference between the standard deviation most
small and the biggest. This can cause problems since the analysis of
variance assumes that the standard deviations of all levels are equal. Select
Variance verification of the list of Tabular Options to execute a test
formal statistics for the difference between the sigmas. Could consider transforming the
values of % methanol alcohol to eliminate any dependency of deviation
mean standard.
ANOVA table for % methanol by Chemists
Source Sum of Gl Square Reason-F Valor-P
Squares Medium
Between 1.04457 3 0.348189 3.25 0.0813
groups
Intra 0.8582 8 0.107275
groups
Total 1.90277 11
(Corr.)

The StatAdvisor
The ANOVA table breaks down the variance of % methanol into two components: a
between-groups component and a within-groups component. The F-ratio, which in this
case is equal to 3.24576, it is the quotient between the estimated between-groups and the estimated
within-groups. Since the p-value of the F-ratio is greater than or equal to 0.05, no
there is a statistically significant difference between the mean of % methanol
between one level of Chemicals and another, with a 95.0% confidence level.

Table of Means for % methanol by Chemists with confidence intervals

of 95.0%
Error Est.
Chemical Case Media (s Limit Limit
s s grouped) Inferior Superior
1 3 84,47 0,189099 84,1617 84.7783
2 3 85,053 0,189099 84,745 85.3617
3
3 3 84,786 0,189099 84,4783 85,095
7
4 3 84,283 0,189099 83,975 84,5917
3
Total 12 84,648
3

The StatAdvisor
This table shows the average % of methanol for each level of Chemicals. Also
show the standard error of each mean, which is a measure of its variability
sampling. The standard error is the result of dividing the standard deviation
jointly among the number of observations at each level. The table also
show an interval around each mean. The intervals currently shown
they are based on Fisher's least significant difference (LSD) procedure.
They are constructed in such a way that if two averages are equal, their intervals will
they will overlap 95.0% of the time. You can view the intervals graphically by selecting
Mean Chart of the Graphics Options list. In the Multiple Range Tests,
these intervals are used to determine which means are significantly different
of others.
Multiple Range Tests for % methyl alcohol by Chemists

Method: 95.0 percent LSD

Chemical Case Media Groups
s Homogeneous
4 3 84,283 X
3
1 3 84.47 XX
3 3 84,786 XX
7
2 3 85,053 X
3

Contrast Sign. Difference +/-

e a Limits
1-2 - 0.616687
0.583333
1-3 - 0,616687
0.316667
1-4 0.186667 0.616687
2-3 0.266667 0.616687
2-4 * 0.77 0.616687
3-4 0.503333 0.616687
* indicates a significant difference.

The StatAdvisor
This table applies a multiple comparison procedure to determine which
Samples are significantly different from others. The lower half of the output shows
the estimated differences between each pair of means. An asterisk has been placed next to
1 pair, indicating that this pair shows statistically significant differences with a
95.0% confidence level. At the top of the page, 2 have been identified.
homogeneous groups according to the alignment of the X's in columns. There are no differences
statistically significant among those levels that share the same column
of X's. The method currently used to discriminate between the means is the
Fisher's Least Significant Difference (LSD) procedure. With this method there is
a 5.0% risk of claiming that each pair of means is significantly different,
when the actual difference is equal to 0.
Multiple Range Tests for % methyl alcohol by Chemists

Method: 95.0 percent Tukey HSD

Chemical Case Media Groups
s Homogeneous
4 3 84,283 X
3
1 3 84.47 X
3 3 84,786 X
7
2 3 85,053 X
3

Contrast Sign. Differentiate +/-

e a Limits
1-2 - 0.856087
0.583333
1-3 - 0.856087
0.316667
1-4 0.186667 0.856087
2-3 0.266667 0.856087
2-4 0.77 0.856087
3-4 0.503333 0.856087
* indicates a significant difference.

The StatAdvisor
This table applies a multiple comparison procedure to determine which ones
the media are significantly different from others. The lower half of the output shows
the estimated differences between each pair of means. There are no statistically significant differences.
significant differences between any pair of means, with a 95.0% confidence level. In the
At the top of the page, a homogeneous group has been identified, according to the alignment.
of the X's in column. There are no statistically significant differences between
those levels that share the same column of X's. The method used
Currently, the procedure for discriminating between the averages is the difference method.
honestly significant difference (HSD) of Tukey. With this method, there is a 5.0% risk of
to say that one or more pairs are significantly different when the actual difference is
equal to 0.
Multiple Range Tests for % methanol by Chemists

Method: 95.0 percent Duncan

Chemical Case Media Groups
s Homogeneous
4 3 84,283 X
3
1 3 84.47 XX
3 3 84,786 XX
7
2 3 85,053 X
3

Contrast Sig. Differenti

e a
1-2 -
0.583333
1-3 -
0.316667
1-4 0.186667
2-3 0.266667
2-4 * 0.77
3-4 0.503333
* indicates a significant difference.

The StatAdvisor
This table applies a multiple comparison procedure to determine which
the means are significantly different from others. The lower half of the output shows
the estimated differences between each pair of means. An asterisk has been placed next to
1 pair, indicating that this pair shows statistically significant differences with a
95.0% confidence level. At the top of the page, 2 have been identified.
homogeneous groups according to the alignment of the X's in columns. There are no differences.
statistically significant between those levels that share the same column
of X's. The method currently employed to differentiate between the means is the
Duncan's multiple comparison procedure. With this method, there is a risk of
5.0% when stating that one or more pairs are significantly different, when the difference
real is equal to 0.
3-14) Three brands of batteries are submitted for study. It is suspected that their lifespans (in
weeks) of the three brands are different. Five batteries of each brand are tested.
with the following results:

Weeks of life
Brand 1 Brand 2 Brand 3
100 76 108
96 80 100
92 75 96
96 84 98
92 82 100

ANSWER
Statistical Summary for %rep. weeks of life
Battery Recount Average Deviation Coefficient of Mínimo Máximo Rang
s o o Standard Variation o
1 5 95.2 3.34664 3.51538% 92.0 100.0 8.0
2 5 79.4 3,84708 4,84518% 75.0 84.0 9.0
3 5 100.4 4.5607 4.54253% 96.0 108.0 12.0
Total 15 91,6667 9,9403 10,844% 75.0 108,0 33,0

Bias Battery Kurtosis

s Standardized Standardized
1 0.46761 -0.27945
2 -0.062529 -1.04148
3 1,36263 1,28626
Total -0.640175 -0.698158

The StatAdvisor
This table shows different statistics of %rep. weeks of life for each one of
the 3 levels of Batteries. The main intention of the one-way analysis of variance is the
to compare the means of the different levels, listed here under the column of
Average. Graphic Selections of Averages from the Graphical Options list to display
graphically the averages.

ANOVA table for %rep. weeks of life by Batteries

Source Sum of Gl Square Reason-F Valor-P
Squares Medium
Between 1196,13 2,598,067 38.34 0,0000
groups
Intra 187.2 12 15.6
groups
Total 1383.33 14
(Corr.)
The StatAdvisor
The ANOVA table breaks down the variance of %rep. weeks of life into two
components: an between-groups component and a within-groups component. The
reason-F, which in this case is equal to 38.3376, is the quotient between the estimated between-
groups and the estimate within-groups. Since the p-value of the F-test is lower
that 0.05, there is a statistically significant difference between the mean of %rep.
weeks of life between one battery level and another, with a 95.0% confidence level.
To determine which means are significantly different from others, select
Multiple Range Tests, from the Table Options list.

Table of Averages for %rep. Weeks of life by Batteries with intervals of

95.0% confidence
Error Est.
Media Case Battery Limit Limit
s s grouped) Inferior Superior
1 5 95.2 1.76635 92.4787 97.9213
2 5 79.4 1.76635 76,6787 82,1213
3 5 100,4 1,76635 97.6787 103,121
Total 15 91,666
7

The StatAdvisor
This table shows the average %rep. weeks of life for each Battery level.
It also shows the standard error of each mean, which is a measure of the
variability of its sampling. The standard error is the result of dividing the deviation
standard pooled among the number of observations at each level. The table
also shows an interval around each mean. The intervals shown
they are currently based on the procedure of the minimum significant difference
(LSD) by Fisher. They are constructed in such a way that if two means are equal, their
intervals will overlap 95.0% of the time. You can see the intervals graphically
selecting Mean Graph from the list of Graph Options. In the Tests of
Multiple Ranges, these intervals are used to determine which averages are
significantly different from others.

Multiple Range Tests for % rep. weeks of life by Batteries

Method: 95.0 percent LSD

Battery Case Media Groups
s s Homogeneous
2 5 79.4 X
1 5 95.2 X
3 5 100.4 X

Contrast Sign. Differentiate +/-

e a Limits
1-2 * 15.8 5.44268
1-3 -5.2 5.44268
2-3 * -21.0 5.44268
* indicates a significant difference.
The StatAdvisor
This table applies a multiple comparison procedure to determine which
averages are significantly different from others. The lower half of the output shows
the estimated differences between each pair of means. The asterisk that is next to
of the 2 pairs indicates that these pairs show statistically significant differences
with a 95.0% confidence level. At the top of the page, there have been
identified 2 homogeneous groups according to the alignment of the X's in columns.
there are statistically significant differences between those levels that share
a single column of X's. The method currently used to discriminate between the
The media is the least significant difference (LSD) procedure of Fisher. With this
Method has a 5.0% risk of stating that each pair of averages is significant.
different, when the real difference is equal to 0.

Method: 95.0 percent Tukey HSD

Battery Case Media Groups
s s Homogeneous
2 5 79.4 X
1 5 95.2 X
3 5 100.4 X

Contrast Sig. Difference +/-

e a Limits
1-2 * 15.8 6.68264
1-3 -5.2 6.68264
2-3 * -21.0 6.68264
* indicates a significant difference.

The StatAdvisor
This table applies a multiple comparison procedure to determine which
means are significantly different from others. The lower half of the output shows
the estimated differences between each pair of means. The asterisk next to it
of the 2 pairs indicates that these pairs show statistically significant differences
with a 95.0% confidence level. At the top of the page, there is
identified 2 homogeneous groups according to the alignment of the X's in columns. No
There are statistically significant differences between those levels that share
a single column of X's. The method currently used to distinguish between the
media is the honest significant difference (HSD) procedure of Tukey.
With this method, there is a 5.0% risk of saying that one or more pairs are
significantly different, when the actual difference is equal to 0.
Method: 95.0 percent Duncan
Battery Case Media Groups
s s Homogeneous
2 5 79.4 X
1 5 95.2 X
3 5 100.4 X

Contrast Sig. Differenti

e a
1-2 * 15.8
1-3 -5.2
2-3 * -21.0
* indicates a significant difference.

The StatAdvisor
This table applies a multiple comparison procedure to determine which
means are significantly different from others. The lower half of the output shows
the estimated differences between each pair of means. The asterisk next to
of the 2 pairs indicates that these pairs show statistically significant differences
with a 95.0% confidence level. At the top of the page, there have been
identified 2 homogeneous groups according to the alignment of the X's in columns. No
there are statistically significant differences among those levels that share
a single column of X's. The method currently used to discriminate between the
means is the multiple comparison procedure of Duncan. With this method, there is a
5.0% risk of saying that one or more pairs are significantly different when the
real difference is equal to 0.