Coding, making the database, and reliability analysis

Having got your questionnaires and tests completed by as many people as you need, you can now enter the data onto a computer database, coding the information where necessary, and ensuring that the records of each person are not named. Then you will work out which are the best, most reliable, items in your scale and discard the rest. You will then have a ‘final’ scale consisting of these reliable items, on which you can work out a final total for each participant and produce descriptive statistics for this final scale (norms). This final scale will then be ready for validation work.

For each stage we describe what has to be done and present a general explanation of how to do it. Appendix 5 provides examples of how to do it in IBM SPSS.

Coding, scoring and data entry

Coding

Responses that do not need coding

• 1. Most answers that are already in numerical form, such as age, and the answers to scale items, if these were given in a numerical form - for example +2, +1, 0, —1, —2, or 5, 4, 3, 2, 1. You do not need to do anything except enter them as they are onto the computer database.
• 2. Answers with a yes/no response format can also be entered directly onto the computer database: 1 for a ‘yes’ and 0 for a ‘no’. Similarly, if you were using a checklist format, answers to each item would be entered as 1 where an item had been checked, and 0 where it had not. (Strictly speaking, this is coding, but it is so simple that you can enter the coded responses straight onto the computer database.)
• 3. Forced-choice format answers, where your participants have to select one of several alternatives. You can enter 1 for items selected, and 0 for those not selected. Later, you can compute scores from this information.

Table 5.1 Examples of items that do not need coding

Table 5.1 shows examples of answers that do not need coding and can be entered directly onto the computer database. If no answer has been given, leave the corresponding space in the database blank, to indicate missing data. There is no need to enter anything special to represent missing information, even though this used to be necessary for older statistics packages. Alphabetical information needs special treatment beyond the scope of this book. Therefore, as just indicated, translate ‘yes’ into T, and ‘no’ into ‘O’.

Information that needs coding

Other information will need to be coded so that you can do statistics. To speed things up and reduce errors, many researchers pre-code their questionnaires by placing the appropriate code next to each response option. Consider the following for example:

How much do you like this chocolate? (tick one answer) (Office use only)

Very much 5

Quite a lot 4

Moderately 3

Not very much 2

Not at all 1

The codes are put in an out-of-the-way column, perhaps with that useful heading ‘Office use only’.

The following are types of information that need coding:

• 1. If your participants answered your scale items by checking one of several ordered responses, instead of writing a number: for example ‘strongly agree’, ‘agree somewhat’, ‘neutral’, ‘disagree somewhat’ or ‘strongly disagree’. In such a case you simply translate the answers into a simple ordered number scale. In this case, it could be +2, +1, 0, —1, —2, or 5, 4, 3, 2, 1. It does not matter which, as long as you consistently apply the same scale to each item. You can use this procedure for a scale with any number of points: 3, 5, 7, 9 and 11 are the most common. You might wonder what to do about ‘reverse-meaning’ items, where agreeing with such items suggests an ‘opposite’ set of beliefs or attitudes or feelings, compared to agreeing with other items. In such cases, you still give the same ‘score’ to ‘strongly agree’ and so forth. However, later, you will have to tell the computer to reverse the scoring of such items. Appendix 5 offers instructions.
• 2. A Visual Analogue Scale (VAS) is a horizontal line across the page, on which the person is asked to place a mark to indicate their feelings or beliefs. Figure 5.1 shows an example. As stated in Chapter 2, responses to these scales are more time consuming to score than other types of response, so consider whether it is worthwhile to use this method.

The example suggests that the person tested perceived the chocolate as sweet and creamy, bland and rather boring. With a VAS, you score the answers by measuring the distance in centimetres or millimetres from the left-hand end of the line. This number is entered onto the database. When describing variables assessed on a VAS, use the word on the right-hand side of the scale. If necessary, reverse the scoring (by subtraction) once all the data are entered in the database.

3. If you have asked for background socio-demographic data from your participants, some of these (such as gender, occupation or group membership) will need to be translated into an appropriate number scale. You should be aware of what type of number scale (interval, ordinal or categorical) you are using because this will affect the kinds of statistics that are appropriate. Consult Appendices 5.1 and 5.2 if you feel doubtful.