One-Way ANOVA

In This Topic

 Step 1: Determine whether the differences between group means are statistically
significant

 Step 2: Examine the group means

 Step 3: Compare the group means

 Step 4: Determine how well the model fits your data

 Step 5: Determine whether your model meets the assumptions of the analysis

Step 1: Determine whether the differences between group means are statistically significant

To determine whether any of the differences between the means are statistically significant,
compare the p-value to your significance level to assess the null hypothesis. The null
hypothesis states that the population means are all equal. Usually, a significance level
(denoted as α or alpha) of 0.05 works well. A significance level of 0.05 indicates a 5% risk of
concluding that a difference exists when there is no actual difference.

P-value ≤ α: The differences between some of the means are statistically significant

If the p-value is less than or equal to the significance level, you reject the null hypothesis and
conclude that not all of population means are equal. Use your specialized knowledge to
determine whether the differences are practically significant. For more information, go
to Statistical and practical significance.

P-value > α: The differences between the means are not statistically significant

If the p-value is greater than the significance level, you do not have enough evidence to reject
the null hypothesis that the population means are all equal. Verify that your test has enough
power to detect a difference that is practically significant. For more information, go
to Increase the power of a hypothesis test.

Analysis of Variance

Source DF Adj SS Adj MS F-Value P-Value

Paint 3 281.698 93.8993 6.02 0.0043

Error 20 312.068 15.6034

Total 23 593.766
Key Result: P-Value

In these results, the null hypothesis states that the mean hardness values of 4 different paints
are equal. Because the p-value is 0.0043, which is less than the significance level of 0.05, you
can reject the null hypothesis and conclude that some of the paints have different means.

Step 2: Examine the group means

Use the interval plot to display the mean and confidence interval for each group.

The interval plots show the following:

 Each dot represents a sample mean.

 Each interval is a 95% confidence interval for the mean of a group. You can be 95%
confident that a group mean is within the group's confidence interval.

IMPORTANT

Interpret these intervals carefully because making multiple comparisons increases the type 1
error rate. That is, when you increase the number of comparisons, you also increase the
probability that at least one comparison will incorrectly conclude that one of the observed
differences is significantly different.

To assess the differences that appear on this plot, use the grouping information table and
other comparisons output (shown in step 3).

In the interval plot, Blend 2 has the lowest mean and Blend 4 has the highest. You cannot
determine from this graph whether any differences are statistically significant. To determine
statistical significance, assess the confidence intervals for the differences of means.
Step 3: Compare the group means

If your one-way ANOVA p-value is less than your significance level, you know that some of
the group means are different, but not which pairs of groups. Use the grouping information
table and tests for differences of means to determine whether the mean difference between
specific pairs of groups are statistically significant and to estimate by how much they are
different.

Grouping Information table

Use the grouping information table to quickly determine whether the mean difference
between any pair of groups is statistically significant.

Groups that do not share a letter are significantly different.

Tests for differences of means

Use the confidence intervals to determine likely ranges for the differences and to determine
whether the differences are practically significant. The table displays a set of confidence
intervals for the difference between pairs of means. The interval plot for differences of means
displays the same information.

Confidence intervals that do not contain zero indicate a mean difference that is statistically
significant.

Depending on the comparison method you chose, the table compares different pairs of groups
and displays one of the following types of confidence intervals.

 Individual confidence level

The percentage of times that a single confidence interval includes the true difference between
one pair of group means, if you repeat the study multiple times.

 Simultaneous confidence level

The percentage of times that a set of confidence intervals includes the true differences for all
group comparisons, if you repeat the study multiple times.

Controlling the simultaneous confidence level is particularly important when you perform
multiple comparisons. If you do not control the simultaneous confidence level, the chance
that at least one confidence interval does not contain the true difference increases with the
number of comparisons.

For more information about how to interpret the results for Hsu's MCB, go to What is Hsu's
multiple comparisons with the best (MCB)?

Grouping Information Using the Tukey Method and 95% Confidence

 Paint N Mean Grouping

Blend 4 6 18.067 A

Blend 1 6 14.733 A B

Blend 3 6 12.983 A B

Blend 2 6 8.567 B

Means that do not share a letter are significantly different.

Key Results: Mean, Grouping

In these results, the table shows that group A contains Blends 1, 3, and 4, and group B
contains Blends 1, 2, and 3. Blends 1 and 3 are in both groups. Differences between means
that share a letter are not statistically significant. Blends 2 and 4 do not share a letter, which
indicates that Blend 4 has a significantly higher mean than Blend 2.

Tukey Simultaneous Tests for Differences of Means

Difference Difference SE of T- Adjusted

of Levels of Means Difference 95% CI Value P-Value

Blend 2- -6.167 2.281 (-12.553, -2.70 0.0606

Blend 1 0.219)

Blend 3- -1.750 2.281 (-8.136, -0.77 0.8682

Blend 1 4.636)

Blend 4- 3.333 2.281 (-3.053, 1.46 0.4779

Blend 1 9.719)

Blend 3- 4.417 2.281 (-1.969, 1.94 0.2450

Blend 2 10.803)

Blend 4- 9.500 2.281 (3.114, 4.17 0.0025

Blend 2 15.886)

Blend 4- 5.083 2.281 (-1.303, 2.23 0.1495

Blend 3 11.469)

Individual confidence level = 98.89%

Key Results: 95% CI, Individual confidence level

In these results, the confidence intervals indicate the following:

 The confidence interval for the difference between the means of Blend 2 and 4 is
3.114 to 15.886. This range does not include zero, which indicates that the difference
is statistically significant.

 The confidence intervals for the remaining pairs of means all include zero, which
indicates that the differences are not statistically significant.

 The 95% simultaneous confidence level indicates that you can be 95% confident that
all the confidence intervals contain the true differences.

 Each individual confidence interval has a confidence level of 98.89%. This result
indicates that you can be 98.89% confident that each individual interval contains the
true difference between a specific pair of group means. The individual confidence
levels for each comparison produce the 95% simultaneous confidence level for all six
comparisons.

Step 4: Determine how well the model fits your data

To determine how well the model fits your data, examine the goodness-of-fit statistics in the
model summary table.

Use S to assess how well the model describes the response.

S is measured in the units of the response variable and represents the how far the data values
fall from the fitted values. The lower the value of S, the better the model describes the
response. However, a low S value by itself does not indicate that the model meets the model
assumptions. You should check the residual plots to verify the assumptions.

R-sq

R2 is the percentage of variation in the response that is explained by the model. The higher
the R2value, the better the model fits your data. R2 is always between 0% and 100%.

A high R2 value does not indicate that the model meets the model assumptions. You should
check the residual plots to verify the assumptions.

R-sq (pred)

Use predicted R2 to determine how well your model predicts the response for new
observations.Models that have larger predicted R2 values have better predictive ability.

A predicted R2 that is substantially less than R2 may indicate that the model is over-fit. An
over-fit model occurs when you add terms for effects that are not important in the population,
although they may appear important in the sample data. The model becomes tailored to the
sample data and therefore, may not be useful for making predictions about the population.

Predicted R2 can also be more useful than adjusted R2 for comparing models because it is
calculated with observations that are not included in the model calculation.

Model Summary

S R-sq R-sq(adj) R-sq(pred)

3.95012 47.44% 39.56% 24.32%

Key Results: S, R-sq, R-sq (pred)

In these results, the factor explains 47.44% of the variation in the response. S indicates that
the standard deviation between the data points and the fitted values is approximately 3.95
units.

Step 5: Determine whether your model meets the assumptions of the analysis

Use the residual plots to help you determine whether the model is adequate and meets the
assumptions of the analysis. If the assumptions are not met, the model may not fit the data
well and you should use caution when you interpret the results.

Residuals versus fits plot

Use the residuals versus fits plot to verify the assumption that the residuals are randomly
distributed and have constant variance. Ideally, the points should fall randomly on both sides
of 0, with no recognizable patterns in the points.

The patterns in the following table may indicate that the model does not meet the model
assumptions.

What the pattern may

Pattern
indicate

Fanning or uneven spreading of residuals Nonconstant variance

across fitted values

A point that is far away from zero An outlier

In this residual versus fits plot, the points appear randomly scattered on the plot. None of the
groups appear to have substantially different variability and no outliers are apparent.

Residuals versus order plot

Use the residuals versus order plot to verify the assumption that the residuals are independent
from one another. Independent residuals show no trends or patterns when displayed in time
order. Patterns in the points may indicate that residuals near each other may be correlated,
and thus, not independent. Ideally, the residuals on the plot should fall randomly around the

center line:

If you see a pattern, investigate the cause. The following types of patterns may indicate that

the residuals are dependent.

Trend
Shift

Cycle

In the residual versus order plot, the residuals fall randomly around the centerline.

Normality plot of the residuals

Use the normal probability plot of residuals to verify the assumption that the residuals are
normally distributed. The normal probability plot of the residuals should approximately
follow a straight line.

The patterns in the following table may indicate that the model does not meet the model
assumptions.

Pattern What the pattern may indicate

Not a straight line Nonnormality

A point that is far away from the line An outlier

Changing slope An unidentified variable

 NOTE

If your one-way ANOVA design meets the guidelines for sample size, the results are not
substantially affected by departures from normality.

In this normal probability plot, the residuals appear to generally follow a straight line. From
the residuals versus fits plot, you can see that there are six observations in each of the four
groups. Because this design does not meet the sample size guidelines, it is important to
satisfy the normality assumption so that the test results are reliable.

Analysis of variance table for Analyze

Taguchi Design
Learn more about Minitab 18

Find definitions and interpretation guidance for every statistic in the Analysis of Variance
table.

In This Topic

 DF
 Seq SS
 Adj SS
 Adj MS
 F-value
 P-Value
DF
The total degrees of freedom (DF) are the amount of information in your data. The
analysis uses that information to estimate the values of unknown population parameters.
The total DF is determined by the number of observations in your experiment. The DF for
a term show how much information that term uses. Increasing your sample size provides
more information about the population, which increases the total DF. Increasing the
number of terms in your model uses more information, which decreases the DF available
to estimate the variability of the parameter estimates.

Seq SS
Sequential sums of squares are measures of variation for different components of the
model. Unlike the adjusted sums of squares, the sequential sums of squares depend on
the order the terms are entered into the model. In the Analysis of Variance table, Minitab
lists the sequential sums of squares for the main effects, interactions, and error term.

Seq SS Term

The sequential sum of squares for a term is the unique portion of the variation
explained by a term that is not explained by the previously entered terms. It
quantifies the amount of variation in the response data that is explained by each
term as it is sequentially added to the model.

Seq SS Error
The error sum of squares is the sum of the squared residuals. It quantifies the
variation in the data that the predictors do not explain.

Seq SS Total
The total sum of squares is the sum of the term sum of squares and the error sum
of squares. It quantifies the total variation in the data.

Interpretation
Minitab uses the adjusted sums of squares to calculate the p-value for a term.
Minitab also uses the sums of squares to calculate the R2 statistic. Usually, you
interpret the p-values and the R2 statistic instead of the sums of squares.

NOTE
In an orthogonal design, the sequential sum of squares is the same as the
adjusted sum of squares.
Adj SS
Adjusted sums of squares are measures of variation for different components of
the model. The order of the predictors in the model does not affect the
calculation of the adjusted sum of squares. In the Analysis of Variance table,
Minitab separates the sums of squares into different components that describe
the variation due to different sources.

Adj SS Term

The adjusted sum of squares for a term is the decrease in the error sum of
squares compared to a model with only the other terms. It quantifies the amount
of variation in the response data that is explained by each term in the model.

Adj SS Term
The adjusted sum of squares for a term is the increase in the regression sum of
squares compared to a model with only the other terms. It quantifies the amount
of variation in the response data that is explained by each term in the model.

Adj SS Error
The error sum of squares is the sum of the squared residuals. It quantifies the
variation in the data that the predictors do not explain.

Adj SS Total
The total sum of squares is the sum of the term sum of squares for an orthogonal
design and the error sum of squares. It quantifies the total variation in the data.

Interpretation
Minitab uses the adjusted sum of squares to calculate the p-value for a
term. Minitab also uses the sums of squares to calculate the R2 statistic.
Usually, you interpret the p-values and the R2 statistic instead of the
sums of squares.

Adj MS
Adjusted mean squares measure how much variation a term or a model
explains, assuming that all other terms are in the model, regardless of
the order they were entered. Unlike the adjusted sums of squares, the
adjusted mean squares consider the degrees of freedom.
The adjusted mean square of the error (also called MSE or s2) is the
variance around the fitted values.

Interpretation
Minitab uses the adjusted mean squares to calculate the p-value for a
term. Minitab also uses the adjusted mean squares to calculate the
adjusted R2 statistic. Usually, you interpret the p-values and the adjusted
R2 statistic instead of the adjusted mean squares.

F-value
The Analysis of Variance table lists an F-value for each term. The F-value
is the test statistic used to determine whether the term is associated
with the response.

Interpretation
Minitab uses the F-value to calculate the p-value, which you use to
make a decision about the statistical significance of the terms. The p-
value is a probability that measures the evidence against the null
hypothesis. Lower probabilities provide stronger evidence against the
null hypothesis.

A sufficiently large F-value indicates that the term or model is

significant.

If you want to use the F-value to determine whether to reject the null
hypothesis, compare the F-value to your critical value. You can calculate
the critical value in Minitab or find the critical value from an F-
distribution table in most statistics books. For more information on
using Minitab to calculate the critical value, go to Using the inverse
cumulative distribution function (ICDF) and click "Use the ICDF to
calculate critical values".

P-Value
The p-value is a probability that measures the evidence against the null
hypothesis. Lower probabilities provide stronger evidence against the
null hypothesis.
 Interpretation
To determine whether the association between the response and each
term in the model is statistically significant, compare the p-value for the
term to your significance level to assess the null hypothesis. The null
hypothesis is that there is no association between the term and the
response.

Usually, a significance level (denoted as α or alpha) of 0.05 works well. A

significance level of 0.05 indicates a 5% risk of concluding that the
coefficient is not 0 when it is. Frequently, a significance level of 0.10 is
used for evaluating terms in a model.

P-value ≤ α: The association is statistically significant

If the p-value is less than or equal to the significance level, you can conclude that
there is a statistically significant association between the response characteristic
and the term.

P-value > α: The association is not statistically significant

If the p-value is greater than the significance level, you cannot conclude that
there is a statistically significant association between the response characteristic
and the term. You may want to refit the model without the term.

If there are multiple predictors without a statistically significant association with

the response, you can reduce the model by removing terms one at a time. For
more information on removing terms from the model, go to Model reduction.