Paper title: Statistical methods for psychological

research-II
Theoretical questions (Unit-III)

Note: The text is taken directly from selected portions from Minimum and King book with
slight modifications
Dr. Deepesh Rathore
Q1. What is one way ANOVA? Why is this test preferred over t-test?

Or

When do we use one way ANOVA? Why is it called an omnibus test?

Answer:

In the 1920s, Sir Ronald Fisher developed a technique called analysis of variance, abbreviated
ANOVA. The technique allows us to compare simultaneously several means with the level of
significance specified by the investigator. one-way analysis of variance is an analysis of variance with
only one independent variable; a single-factor design.

In one-way analysis of variance, there may be two or more treatment conditions, often referred to as
different levels of the categorical independent variable or factor. We will use k to represent the
number of treatment conditions. The k treatments may be identified by letters, such as A, B, and C,
and the population means as 𝜇𝐴 , 𝜇𝐵 , 𝜇𝐶 . If the different treatments have no effect on the variable
under observation, then we expect these population means to be equal. We test the null hypothesis:

𝐻0 = 𝜇𝐴 = 𝜇𝐵 = 𝜇𝐶
The null hypothesis in ANOVA is often referred to as an omnibus hypothesis (i.e., covering many
situations at once), and ANOVA itself as an omnibus test. The alternative hypothesis, often stated
simply as “not H0,” is that the population means are different in some way. For example, two or more
group means may be alike and the remainder different, all may be different, and so on.

When we have more than two groups we prefer ANOVA over t-test because t-test takes only two
means (two groups) at a time and hence in case of three groups, A, B, and C we would conducting
multiple t-tests between, A and B, B and C, and A and C groups. Also, with every t-test there is a type
I error associated with it so multiple t-test would mean higher chances of committing at least one
type I error.

Q2. What is the basis of one-way analysis of variance?

Or

What are the various sources of errors associated with one way ANOVA?

Answer:

The total variation among all the scores and divide it (statisticians say “partition” it) into two parts:
 1. Within-groups variation
2. Between-groups variation

1. Within-groups variation: Within each sample, the individual scores vary around the sample
mean. We will call this within-groups variation. Remember what you have learned about the
influence of random sampling variation. Within-groups variation is a direct reflection of the
inherent variation (variation due to chance) among individuals given the same treatment.
Within-groups variation reflects only inherent variation. It does not reflect differences
caused by differential treatment.

2. Between-groups variation: is also a reflection of the inherent (chance) variation among

Q3. What is F ratio? Why it is always greater than 0?

Or
What is F ratio? under what conditions it will be greater then or less than 1?

Or

What do you understand by F ratio? Why is the region of rejection in ANOVA is on the upper tail of
F distribution?

Answer:

To complete the analysis of variance, we require a method of comparing between group variation
with within group variation. This is the F ratio, a statistic named in honor of Ronald Fisher. The F ratio
is formed by the ratio of the two independent variance estimates:

2 2
In constructing this ratio, 𝑠𝑏𝑒𝑡 is always placed in the numerator and 𝑠𝑤 is always placed in the
2 2
denominator. If H0 is true and there is no treatment effect, 𝑠𝑏𝑒𝑡 and 𝑠𝑤 will both be estimates of the
inherent variance only, and the ratio should be about 1.0.
𝐼𝑛ℎ𝑒𝑟𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑛𝑐𝑒 + 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 𝑒𝑓𝑓𝑒𝑐𝑡
𝐹=
𝐼𝑛ℎ𝑒𝑟𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑛𝑐𝑒
2
If H0 is false, 𝑠𝑏𝑒𝑡 will estimate the inherent variance plus the treatment effect and the ratio should
exceed 1.0. The greater the treatment effect, the more F tends to exceed unity.

When H0 is true and certain assumptions hold (see Section 18.11), F ratios follow the theoretical F
distribution. Like the t distribution, the F distribution is actually a family of curves depending on the
degrees of freedom.

If the estimate of 𝜎 2 in the numerator is smaller than that in the denominator, F will be less than
1.00 but not less than zero (because variance estimates are never negative). But if the estimate in
the numerator is larger than that in the denominator, the F ratio may be much larger than 1.00. The
null hypothesis of equality of population means will be rejected only if the calculated value of F is
larger than that expected through random sampling if the hypothesis is true. Consequently, the
region of rejection is entirely in the upper tail of the F distribution

Q4. What is the relationship between F and t statistic?

Answer:

ANOVA is closely related to the t test and, for the special case of two groups, leads to exactly the
same conclusions as the t test. In fact, you can use ANOVA instead of t in the two-sample,
independent-groups design.
If we know the value of t statistic, we can calculate the value of F statistic and vice versa.

For example: if t=4, then what is the value of F?

Since, F= t2, F=42=16

Another way: if F=9, then what is the value of t?

Since, F= t2, t=√𝐹 = ±3

The above relationship is applicable only for two independent samples.

Q5. What are the assumptions associated with one way ANOVA? Explain the cases of moderate
departure from these assumptions.

Answer:

Because F= t2, you should not be surprised to learn that the assumptions underlying the F test for
one-way ANOVA are the same as those for the t test for independent samples:

1. The populations are normally distributed.

2. The variances of the several populations are the same (homogeneity of variance).

3. Selection of elements comprising any particular sample is independent of selection of elements of

4. Samples are drawn at random with replacement.

As with the t test for independent means, moderate departure from the normal distribution
specified in the first assumption does not unduly disturb the outcome of the test. This is especially
true as sample size increases.

Violations of the second assumption, homogeneity of variance, are also negligible when sample sizes
2 2
are equal. Heterogeneity of variance ordinarily becomes a problem only when 𝜎𝑙𝑎𝑟𝑔𝑒𝑠𝑡 = 4𝜎𝑠𝑚𝑎𝑙𝑙𝑒𝑠𝑡
or, with more moderate differences, when sample sizes differ considerably.

The third assumption, independence of samples, reminds us that the ANOVA procedure described to
this point is not appropriate for repeated measures on the same subjects nor when matched subjects
have been assigned to treatment groups.

The biggest problem is probably that of obtaining random samples. We used random sampling, but
in the real world most researchers use random assignment of available subjects. Random sampling is
more of a premise than a requirement—it allows us to extend conclusions to the population.

Q6. What is a post-hoc test? When it is used?

Or

what do you understand by post-hoc comparisons?

If the obtained F ratio turned out to be smaller than the critical value. We would conclude that our
data do not support the presence of treatment effects, and that would be that. There would be no
justification for further statistical work.

However, when obtained F ratio is greater than critical F, it will led us to reject 𝐻0 = 𝜇𝐴 = 𝜇𝐵 = 𝜇𝐶
in favor of the very broad alternative hypothesis (not H0) that the three population means differed in
some way. But where is the real difference (or differences)? All three might differ from one another,
or two of them might be equal and differ from the third. To answer this question we may, on
obtaining a significant overall F, proceed with further statistical comparisons involving the group
means. The tests we use to make these specific comparisons are called post hoc comparisons or a
posteriori (Latin, meaning “what comes later”) comparisons.

If we are comparing many means, we would expect some of the differences to be fairly substantial as
a result of random sampling variation alone, even if the null hypothesis is true. Applying the
conventional t test just to the largest or to all of the possible differences between sample means
would substantially increase the probability of making a Type I error—that is, rejecting H0 when it is
true. Instead of using a sampling distribution that compares the means of only two samples (as is
done with t), post hoc tests generally employ sampling distributions that compare the means of
many samples.

In short, post hoc tests protect us from making too many Type I errors by requiring a bigger
difference (between sample means) before we can declare that difference to be statistically
significant.

There are several commonly used post hoc tests. The only real difference among them is that some
are more conservative than others. In ascending level of conservativeness with regard to Type I
errors (but descending power), we could choose from among such tests as Duncan’s multiple-range
test, the Newman–Keuls test, Tukey’s HSD test, and the Scheffé test. To use any one of these tests,
our F ratio must first be significant.