Univariate analysis

In this section, you will learn how to create a new variable, run a frequency statistic, calculate the mean of a continuous variable, and build a histogram.

Before we begin, please download the Crime Survey for England Wales 2011-2012 (or CSEW) from the UK Data Service website. This way, you'll be able to follow along with our analyses.

Instructions on how to locate and download the CSEW are available here.

What factors influence police confidence?

Univariate analysis refers to the quantitative data exploration we do at the beginning of any analysis. These analyses provide us with descriptions of single variables we are interested in using in more advanced tests and help us narrow down exactly what types of bivariate and multivariate analyses we should carry out.

We’ll start our univariate analysis with operationalizing our research question. We can do this by determining which variable most suits our concept of confidence in the police. There is no one variable in the CSEW dataset that concerns police confidence. There are, however, six variables, called polatt1 through polatt7, which address the CSEW respondents’ opinions of how successful the police are at curtailing crime and how much trust the respondents have in their local police force.

Each of these variables contains answers to questions about confidence in the police, and because we are curious about all aspects of police confidence, we are interested in the responses to all of these questions. We could analyse each of them separately, but overall confidence is a good place to start. Because we want to analyse the responses to all of the polatt variables, we can use SPSS to combine these six questions about police confidence together, creating one new variable for us to use.

(Important note: we are only creating a new variable because we don’t have one better suited for us in our dataset. Transforming variables is done when you need the data to work better for you. This is not necessarily a technique that needs to be employed every time you select a dependent or independent variable. However, knowing how to transform variables is really useful when you find yourself with a research question that requires some tailoring. This website hopes to show you how to troubleshoot certain issues that may arise when you’re working with real, full datasets. One of these potential issues is the lack of a functional, pertinent dependent variable, and this section will show you how to overcome this problem through variable transformation.)

Polatt1-polatt7 are categorical variables, with responses falling into distinct categories on a 1 through 5 scale. When we combine all of the polatt variables into one single variable, we’ll have created a continuous variable. This is because SPSS will collapse each respondent’s six polatt responses into one total police confidence score, giving us a range of numerical scores from 6 to 30. For example, a respondent who answered "1" or "Strongly agree" in response to all six of the polatt questions would score a 6 on our new confidence in the police scale, as 1+1+1+1+1+1 = 6. SPSS will do this calculation for each respondent's answers to the polatt questions, and provide us with a continuous range of values we can use in our analyses.

There are a few necessary conditions to keep in mind if you decide to transform categorical variables into a continuous variable:

• The categorical variables must have data collected from the whole sample. In some datasets, like the CSEW, some variables represent the responses to questions only some of the survey participants have answered, because some questions are only asked of part of the survey sample. You can check to make sure your variables meet this condition by searching for them in the survey questionnaire and ensuring that they have been asked of all participants.
• The categorical variables you select need to contain data concerning the same basic questions. For example, you would not want to compress data from a variable about the number of dogs in a neighbourhood with a variable about the amount of money made selling cars. Those two variables use two different measures to answer two different questions.
• The categorical variables must use the same units of measure. In our case, the categorical variables we select must all use the same point scale, either “1 through 4” or “1=strongly agree to 5=strongly disagree.” In addition, these scales need to be going in the same direction, meaning that you wouldn’t collapse two variables together if one was measured with “1=strongly agree” and the other was measured with “1=strongly disagree.” (If they do not use the same point scale, or use scales going in different directions, we can transform the data in the variables so that the measurements match.)

Using those conditions, it appears that the variables polatt1-polatt7 are acceptable categorical variables to transform into a new continuous variable.

a. The variables polatt1-7 were asked of all the survey respondents

b. Polatt1-7 revolve around the same basic question, concerning confidence in the police. (This information is available in the Label cells of the polatt variable rows in the Variable View window of the CSEW dataset.)

c. Polatt1-7 are measured with the same 5-point scale, heading in the same direction, making the respondent answers easy to collapse together. (You can check these values by clicking on the Values cell in each polatt row in Variable View.)

Now that we know that our selected variables are appropriate, we can begin our univariate analysis! We will be running frequency analyses and creating graphs to check the distribution of the values in our dependent variable.