Basic Applied Techniques
Source: www.spss.com
Copyright SPSS, Inc. 2004


Choose the right stat to make better decisions

The information age has changed the way many of us do our jobs. In the 1980s, we focused on collecting data. By the early 1990s, most organizations were swimming in it. Today, we must analyze this wealth of data and turn it into useful information. Spreadsheets are a start, but they only go so far. Using a statistical package allows you to go beyond basic row-and-column math and simple summaries for better decision-making. In this paper, we will discuss how statistical techniques help you get more useful information out of your data.


Add value to decision-making
Using statistics can help analyze data to make better informed decisions because:
· statistics are useful – they summarize data or graphically display relationships and quickly identify unusual points;
· statistics are easy to understand – they are easy to read and interpret with straightforward formats, tables and charts; and
· statistics add value to decision-making – they allow you to quantify relationships (instead of relying on hunches) and go beyond information spreadsheets and databases provide.

By understanding basic statistical techniques, you can learn more about your data, discover interesting relationships, better use scarce resources and give credibility to your ideas.


Measure what you want to analyze
Data is a measurement of some occurrence or event. By determining your data type, you can select a suitable statistical procedure. On the flip side, to analyze data a certain way, you must collect it a certain way.

For example, if you want to measure work experience, you have a few choices. You can ask a respondent for a specific number of years (two years or five years). Or you can ask each respondent for a category (less than two years, two to five years). If your supervisor wants a graph showing how many employees are at each experience level, you could present information in a bar chart. You would choose a bar chart because it displays data collected in categories. However, if your data reflected a specific number of years for each respondent, you would probably choose another type of chart or graph.

Levels of measurement explain how data are measured: categorically or continuously. Categorical data are measured in a group (less than two years, two to five years). Continuous data are measured with specific numbers (two years or five years). In addition, categorical data can have two distinct variables: nominal and ordinal. Nominal variables are collected in categories of group identifiers (department, business unit). Ordinal variables have implied scales (satisfaction levels, education level).

There are also two types of continuous variables: interval and ratio. Interval variables have a numeric scale without a true zero (temperature, achievement test scores). Ratio variables have a measurable scale and a true zero designating the absence of what is measured (age, net worth). With both interval and ratio variables, a one-unit increase anywhere on the scale represents the same change in quantity. Regardless of whether variables are nominal, ordinal, interval or ratio, they can help you learn more about your data.


When to use statistical techniques
Now that we defined levels of measurement, variable types and how they affect data analysis, let’s discuss when and how to use specific statistical techniques.

Nominal: categories with group identifiers (gender, region, country)
Ordinal: categories with implied scales (military rank, satisfaction levels)
Interval: variables in a numeric scale without a true zero (temperature, test scores)
Ratio: variables in a measurable scale with a true zero (annual income, vacation)

Bar charts. A graphic overview of a categorical variable shows easy-to-understand information. Like frequency distribution, a bar chart offers a range of answers while identifying the most frequent response. 

Crosstabulation tables. With crosstabulation tables, or crosstabs, you can look at the frequencies of two variables at once. Crosstabs help you determine whether variables are related, and if so, can measure the strength of that relationship. And like frequencies, you can use crosstabs to examine for data entry errors. The number of distinct categories in each variable determine the table’s size, with cells created at the intersection of each row and-column combination.


Techniques for continuous data
Descriptives. Descriptive statistics give you a quick overview of continuous variables by calculating the average, spread, minimum, maximum and number of cases for each variable. These statistics work like frequencies for categorical data and can help identify data entry errors. They are especially useful when seeing a dataset for the first time. For example, you may want to know the average age, education level, beginning salary, current salary and work experience of employees in your organization. Descriptives summarize this information, helping you quickly learn more about your data as a whole.

Histogram. In our earlier discussion of levels of measurement, we considered the work experience of bank employees when measured in categories of less than two years, two to five years, etc. But instead of looking at categories, you may want to see work experience measured on a continuous scale (a specific number of years). Used with continuous data, a histogram provides an overview of distributed data values.

Boxplot. Again, rather than getting your data overview in table form, you can draw similar conclusions from a graph. A boxplot is a good method for identifying unusual values and gaining insight into the pattern of a majority of values. It shows more than a spreadsheet, displaying the minimum, maximum, range, average and outliers, or extreme cases. The boxplot also shows data distribution. 

Scatterplot. Rather than limiting your investigation to individual variables, you may want to examine whether two continuous variables are related. Plotting the intersecting points on axes can quickly isolate trouble spots and identify groupings. Scatterplots draw attention to unusual points that may affect averages. These graphs also provide a good overview of the relationship between variables.