Let's start with a brief review. In the last
several labs we looked at ways to test samples.
We used z-scores to compare a treatment sample
with a population for which we knew the
population mean μ and the population standard
deviation s
We used one-sample t-tests to examine the
same situation, except we don't know the
population s, so
we need to use an estimate (the sample
standard deviation).
The last lab used a different computational
formula to calculate the observed t for two
more situations:
- repeated measures, in which there is one
sample, but each individual is tested twice.
- matched pairs, in which there are two
samples, but they are related on a subject
by subject basis.
The logic of today's lab should seem similar to
the last few. The overall logic is the same, we
still use the t-distribution to find our
critical values. However, things get a little
more complicated, because of the situation that
we are interested in. Now we are going to look
at a situation where we are interested in the
potential difference between two different
populations, where each population is
represented by a separate (and unrelated)
sample. And again, we'll deal with situations in
which we don't know the μ or s for either of these
populations, so we'll have to use samples to
estimate them.
An experiment that
uses a separate independent sample for
each treatment condition (or each
population) is called an independent-measures
research design. Often you'll also see
it referred to as a between-subjects
or between-groups
design.
|
So we'll use the same logic and steps for
hypothesis testing that we used in the previous
labs, and fill in the details of the differences
as we go.
Step 1: State your H0 and
H1 and choose your criterion: α
Step 2: Collect the
sample
Step 3: Compute the
tobs for your sample
Step 4: Compare tobs
to tcrit (or p to α) and make a decision
Let's start with Step 1:
Figuring out your critera is exactly the
same process as before, you pick what your
field has decided as being an accepted level
of alpha (chance of making a type I error).
For our example, let's assume a
= 0.05
The hypotheses are going
to be a bit different, because the
situation is different. Remember, that now
we are making hypotheses about two
different populations, not just comparing
a treatment to what is known.
For example, suppose
that you want to compare two different
treatments (e.g., two ways of studing, two
different drugs, etc), or you want to
compare two groups of people (e.g., men
vs. women, young vs. old, etc.). So now,
the hypotheses are about population A
(men) and population B (women), and how
they are different from one another.
Is this going to be a one-tailed test
or a two-tailed test? In this case,
we'll conduct a two-tailed test. We
won't make a directional prediction.
So the H0 hypothesis would
be that men and women are the same
height. That is,
H0: μMen = μWomen
- or -
H0: μMen - μWomen
= 0
Our alternative hypthesis would be
that men and women are different mean
heights. That is,
H1: μMen not
equal to mWomen
- or -
H1: μMen - μWomen
not equal to 0
Note: What might the
hypothesis be for a one-tailed
test? Men are taller than women.
H0: μMen<
μWomen
H1: μMen
> μWomen
|
Step 2 : Criterion
for decision: α= .05
Step 3 & 4: Sample
& test statistics
We are going to be using two
samples, one to represent each population.
Men's heights in inches: 67,
73, 74, 70, 70, 75, 73, 68, 69
Women's heights in inches: 69, 63, 67, 64, 61,
66, 60, 63, 63
Remember that because we're
using samples, we can only estimate the values
of the population parameters, and so we need
to take degrees of freedom into account. We've
got two samples. How many values are free to
vary?
Now comes what will look to
be the big difference. We need to compute
our observed t statistic. Basically, at the
conceptual level, the formula is the same.
However, at the practical level, it is more
complex because we have two samples, which
means that we have two estimates.
tobs =
The numerator is
straightforward:
= the difference between
the two sample means
(μA - μB) = 0:
this is H0, and that's what
we're testing
This is the an estimate
of the error from the two samples.
Recall that each sample will have some
sampling error associated with it.
What we need to do here is pool
the error from the two samples. The
reason that we want to pool the
samples is to make the estimate
of the standard error better.
Basically, what we're doing is
increasing the sample size that our
estimate is based on, which will
increase the precision of the
estimate.
- because each sample may be of
different sizes (n's) we need
to weight each sample's estimate of
variability by its degrees of freedom.
Pooled variance =
pooled variance = but we can
simplify this as:
The formula for
our estimated standard error is:
So let's fill in the
numbers from our example. To calculate
by hand, we need to go back to the raw
numbers and compute the SS's and the
sample means. Here are the results or
those computations:
= 71.0, = 64.0
SSA =
64.0, SSB = 66.0
sA = 2.83, sB =
2.87
tobs =
= = 5.22
Step 5 : Compare observed to critical
value and make a decision about the null
hypothesis
What is our critical t?
Go to the table and look
up the value for: 2-tailed, α = 0.05, df
= 16.
So now we compare the two t
statistics.
tobs = 5.22
tcrit = 2.12
Our observed (computed) t
statistic is much greater than the
critical t statistic; in fact, it is
greater than the largest critical t
reported in the t-table for df = 16, p =
.01, 2-tailed. Click on the table above to
find that value. So we feel confident in
rejecting the H0. There does
seem to be a difference between the
heights of men and women.
(1)
A psychologist is interested in
studying the effects of fatigue on
mental alertness. She decides to
study this question using an
independent samples design. She
randomly assigns individuals to
two groups.
After this period, each subject
is tested to see how well they
detect a light on screen (the
dependent variable is the
subjects' number of mistakes which
reflects their mental alertness.
So the higher the number, the less
alert they are.).
Here are the results from the
two groups:
Using an independent samples
t-test, answer the question of
whether fatigue adversely affects
mental alertness (α =
0.05).
Using SPSS to compute independent-samples
t-tests
The data entry for this
test is different from that for related
samples. There we had two measures taken on
the same or related persons. Here we have only
one dependent measure ("readingbefore" in the
screenshots below) taken on two independent
groups (two clubs "Latin" and "Sports"). There
we had two columns for measures; here we have
only one. But as a second column we need to
identify which group the person belonged to.
So we code group membership just as we could
code club by indicating 1 = latin and 2 =
sports.
The usual
practice is to indicate the
independent variable (group) in the
first column and the dependent
variable in the second column. |
|
Go to Analyze and
Compare
Means. The one
that we're interested in today is Independent-
Samples T Test.
|
|
In the next window,
select the variables that you are
testing and then enter your
independent variable into Grouping
Variable.
|
|
Click the button to
Define
Groups
and enter the values that
you used to define the
groups (e.g., 1 for group 1
and 2 for group
2). Then click Continue, and when
you reutrn to the previous
page, OK. |
|
|
Here is what the
output will look like.
|
|
The output includes
statistics for each group, tobs,
degrees of freedom, p-value
(Sig. 2-tailed), the mean difference
(sample mean1 - sample
mean2), and SE of
differences.
Note that SPSS
doesn't tell you to reject or fail
to reject the H0, nor
does it give you the tcrit.
To make your decision about the H0
you must compare the 2-tailed
p-value with your α-level. If the
p-value (or 1/2 of it for a
1-tailed test) is equal to or
smaller than the your α-level, then
you should reject the H0;
otherwise you should fail to reject
H0.
|
Notice that there are
two rows of numbers in the t-test
output. These correspond to the two
t-tests we ran in Excel. Equal
variances are assumed in row 1and not
assumed in row 2. The results are
presented for Levene's Test for
Equality (or homogeneity) of Variance
. If the Levene's test is not
significant (Sig. > 0.05), then we
can assume equal variances and use the
values in row 1. If Levene's test is
significant (Sig. < 0.05), we must
use the values in row 2, which makes
adjustments to more accurately
estimate the t-value in such a case. |
(2):
A psychology instructor at a large university
teaches statistics. Because there are 22
students in the class, he has broken them into
2 groups. Each group has a different graduate
assistant who is responsible for running
separate breakout lecture and lab sections of
the course. One GA has lots of experience
teaching, while the other has more limited
experience. The instructor wants to check for
comparable learning across the two GAs, hoping
to find no difference. The data below are the
scores (out of 100) of the students on the
first midterm. Assume an a
= 0.05 level. The data are as follows (notice
that one group has more students than the
other):
|
Group 1 (less
experienced GA) |
|
Group 2 (more experienced
GA)
|
1 |
60 |
11 |
70 |
2 |
65 |
12 |
85 |
3 |
69 |
13 |
72 |
4 |
58 |
14 |
83 |
5 |
57 |
15 |
81 |
6 |
59 |
16 |
69 |
7 |
52 |
17 |
65 |
8 |
72 |
18 |
75 |
9 |
70 |
19 |
79 |
10 |
65 |
20 |
71 |
|
|
21 |
89 |
|
|
22 |
80
|
Enter
the data into SPSS. Test your H0
using an independent-samples t-test. Create a
bar graph to show the results.
Lab Exercises 3 - 5 show how
different designs affect the following set of
data. Both tests will be to find any difference
between treatments with α = .05.
(These data can represent either 10
different participants or 5 participants
tested in each condition.)
Treatment 1 Treatment 2
10 11
2 5
1 2
15 18
7 9
(3) Assume
that the data are from an independent
samples experiment using two separate
samples, each with 5 subjects. Use SPSS
to test whether the data indicate a
significant difference between the two
treatments (assume a
= .05). List each step of the hypothesis
testing procedure.
(4) Now assume that the data are from
a repeated-measures design using one
sample of 5 subjects, each of whom have
been tested twice. Use SPSS to test
whether the data indicate a significant
difference between the two treatments
(again assume a
= .05). Remember that you'll have to
change the way the data are entered in
the data window.
(5) You should find that the repeated
measures design and the independent
samples design reaches a different
conclusion. How do you explain the
differences (hint: think about how
sampling error is estimated for the two
tests).
|