How to Construct Variables to Make Analysis Easier

Introduction

In addition to the basic workflow for checking and cleaning a project , it is also useful to create certain types of variables up-front that will facilitate additional analysis This article describes methods for constructing variables to make your analysis easier.

Pick One - Multi (e.g., likert scales, importance ratings)

Questions which contain ratings (e.g., ratings of importance, agreement, satisfaction) should be initially setup in Q as Pick One - Multi questions. It is often useful to have a number of additional variants of the data, such as numeric variables for computing means, and variables containing Top 2 Box score with separate versions of each individual item in the scale.

The most straightforward way to create such variables is typically as follows:

From the toolbar menu, select Automate > Browse Online Library > Create New Variables and select one of the following options. Commonly used options are:

• Flatten Question(s) - flattens a grid into a single column of numbers. 
• Top 2 Category Variables(s) - Top 2 Box
• Recode Net Promoter Score (NPS) Variable(s)
• Numeric Variable(s) from Code/Category Midpoints
• Case-Level Shares

Number - Multi (sets of related numeric variables, such as consumption by brand)

Where Number - Multi questions have a small number of unique values, it is often useful to copy them and also represent them in a way that makes the categories visible. This can be done by:

• Select the variables on the Variables and Questions tab. 
• Change the Question Type setting to Pick One - Multi.

Ranking questions

It is often useful to represent ranking data in multiple ways (e.g., percentages, top 3 ranks, means). See How to Analyze Ranking Data for detail on how to achieve this.

Grid questions

It is often useful to have two different version of Pick Any - Grid and Number - Grid questions: one version where the missing values are set at their defaults (i.e., people have missing values for options not seen) and another where the missing value are unchecked (if and as appropriate). For example, if the grid question shows occasions that different brands are consumed and an earlier question in the study was used to filter which brands were shown in the grid question, it would be appropriate to un-check the missing values.

Coding text data

Importing coded data

Where coding has been done in another program, the best approach is usually to have it added to the data file. If you have the data in a different data file, the best approach is:

1. Get both data files as SPSS .sav data files. Note that if you have the data as a CSV or Excel file, you can convert it to an SPSS file using Tools > Save Data as SPSS/CSV File.
2. Merge the data using Tools > Merge Data Files > Add New Variables (see Merge Data Add Variables).
3. Import the data into Q.

Often, coded data is not entered in the standard format for Pick Any data, so it is necessary to set its Question Type to Pick Any - Compact. 

Performing coding within Q

To perform coding within Q, see Coding.

Next

How To Code A Subset Of Responses

How to Scale Respondents to have a Mean of 0 and a Standard Deviation of 1

How to Create Binary Variables

How to Compute Top 2 Box Scores