Week 11: SPSS Pre-Class Worksheet

Introduction

This week we are looking at inference, statistically singnificance, and hypothesis testing. In
particular in this worksheet we learn how to:

• Test the statistical significance of associations observed in a cross-tabulation using

There are 3 discrete tasks to complete below, and some questions to answer along the way.

Please, take your time, repeat steps if necessary, and bring any questions you have to your
workshop.

Please NOTE: This worksheet assumes you have completed previous weeks’ worksheets,
and thus assumes you know how to do certain things already. If you have not completed
the previous worksheets then you should do so before attempting this worksheet.

1
TASK ONE: Using Chi-Square to test the statistical significance of an association between two
categorical variables

1. Download and open the ‘Taking Part 201819.sav’ file from Canvas (the ‘Taking Part’ survey
examines people’s involvement in various cultural and sporting activities in the UK).

We are going to use data contined in this dataset to test the hypothesis that ‘There is a relationship
between someone’s age and the likelihood that they will spend their free time gardening.’

To do so, we will use the variables ‘freetime9’, which tells us whether cases in the sample spend
their freetime gardening or not, and ‘agelong’, which tells us about the age of these cases in terms of
age categories.

2. produce an appropriate univariate analysis of each of the two variables

Remember: before analysis we would want to weight the data, and before working with any
specific variable, we would want to check categories/values have been set as missing
appropriately as has the level of measurement.

3. Produce a cross-tabulation of the two variables, placing the depedent variable in the rows, asking
for column percentatges, and for a Cramer’s V statistic as a measure of the relative strength of
association between the variables.

Q1: What does our anlaysis of the relationship between these variables suggest? Are they
associated? How? Is it a strong or weak association?

Now we will ask SPSS to apply the ‘chi-sqaure’ test to this crosstabulation, the results of which will
tell us whether the association between these variables is statistically significant or not i.e. if there is
enough evidence of an association between these variables to be confident that they actually are
associated. If our Chi-Square suggests that the association is statistically significant, then we can
reject out ‘null’ hypothesis that these two variables are not associated (as very unlkley to be the
case), and thus accept our hypothesis that they are related.

4. To produce the chi-square statistic, ask SPSS to produce the cross tab again, but this time select
‘Statistics’ and select ‘Chi-square’:

2
5. Click ‘continue’ and then ‘OK’ in the cross-tab window to produce your cross-tab as before, tis
time with the results of a chi-square test also.

In your output viewer, you will see your cross-tab again, and also this new output:

The two crucial pieces of information are highlighted above.

Firstly, the note under the table of results tells us whether or not certain important test criteria have
been met. If any more than 20% of cells in our crosstab have expected counts of less than 5, then
the results of this test will be invalid.

In this example, we are being told that 0% of cells have expected counts of less than 5, so our results
can be accepted as valid.

Secondly, under ‘Asymptotic Significance’ we are being given information about whether any
observed association is statistically significant, and at what level.

For an association in a cross-tab to be considered statististically significant, we want this figure to be

In this example, the sig. value is ‘<.001’, which means less than 0.001., thus, the association between
our two variables can be considered statistically significant.

On that basis, we can reject our null hypothesis of no association and accept our hypothesis that
they are associated .

TASK TWO: Testing the statistical significance of a layered cross-tabulation

Last week we learnt how to add another variable into an analysis o the asssociaiton between two
categorical variables, by adding that variable as a set of layers in a cross-tab.

1. Produce the cross-tab between ‘freetime9’ and ‘agelong’ again, but this time add in ‘sex1_3cat’
(sex of respondent)’ as a ‘Layer’ variable, and again ask for the chi-square test results as well as
the cramer’s v statistic.

3
 Q2: Is the association between ‘freetime9’ and ‘agelong’ statistically significant among
each category of respondent sex?

TASK THREE: Producing a confidence interval around a mean, and comparing means between
groups

A ‘confidence interval’ is a range of values around a sample statistic, within which we think the
population statistic will lie.

The Taking Part dataset used in the tasks above has a variable which tells us how many arts
activities participants have participated in or attended in the last year - ‘ArtAct1’

1. Conduct a univariate anlaysis of this variable in SPSS (rembering to check values or level of
measurement)

Q3: What is the mean number of arts activities attended in the last year amongst cases in
this dataset?

We can ask SPSS to produce a confidence interval around this mean.

2. Select ‘Analyze’ from the top menu, then ‘Descriptive Statistics’, then ‘Explore’:

3. In the ‘Explore’ window, move the ‘ArtAct1’ variable into into the ‘Dependent List’ panel and
select ‘OK’:

4
The output produced shows a lower boundary and upper boundary for the 95% confidence interval
around the mean.

Andy Field explains the space in between these boundaries this way: “if we had collected 100
random samples in an identical way, and in each of them calculated the mean and a confidence
interval for that mean, then for 95 of these samples the confidence interval contains the true value of
the mean in the population”.

So, we can be 95% confident that the population statistic is somewhere between 2.88 and 3.01 arts
activities participated in or attended in the last 12 months.

Using the ‘Explore’ function, we can also add in a ‘factor’ variable, which will calculate the same
statistics, but this time for cases in each of the categories of our factor variable. We will use fhte
example of respondent sex

4. Open the ‘Explore’ menu again, and add in ‘sex1_3cat’ into the ‘Factor List’ panel.

This gives us a mean and a confidence interval for the different categories of respondent sex in our
dataset.

Q12: What are the mean and the 95% confidence intervals for ‘males’, ‘females’, and those
in the ‘other’ category? What does comparing the mean and confidence intervals between
these three groups tell us?

END.