Checking the database

Note: At any time, you can click on Edit at the top of the SPSS screen, and choose Options. This will allow you to choose how your results are displayed - for example the type of table your results are displayed in, and whether the variables appear in alphabetical order, or their order in the database. It is not urgent or even essential to change the display type, you can just keep going with the default option. But you might want to experiment with the display options, or use the relevant section of Pallant’s manual.

Before starting statistical analysis, you should carry out the following checks for accuracy. We have described three checks, starting with the quickest and easiest, but if you carry out the second check, there is no real need for the first, and if you carry out the third check, there is no real need for the first two. If you have labelled your values, and these labels are showing up, rather than the numbers you need to check, place the cursor anywhere in the column concerned, and click on the little icon that looks like a luggage label, near the end on the right.

• • First, a preliminary scan (‘eyeballing’). This may only enable you to detect errors that would be detected anyway, in the second and third checks. But it is always a good idea to ‘eyeball’ your data. It is embarrassing to find out - when it is too late to do anything about it - that there are errors that could have been detected if you had just glanced over your material. So, glance over your database. It is usually easier to do this if the database has been printed out. To print, with the database on the screen, either click the Print icon, or select File, then Print. Look quickly over the database for things that should not be there, for example ages like 3025, or values like 53, when you should have typed ‘30’ and ‘25’, ‘5’ and ‘3’ in separate cells. Note any errors, and check your original data for the correct entry. Then go back to the computer database and correct the entries. To correct errors on the computer database, move the cursor to the cell concerned, and type in the correct entry. This will not appear in the cell itself, but in the dialogue box immediately under the tool bars. When you press ‘Enter’, the corrected entry should thenappear in the highlighted cell. Save the new, corrected version of the database, including at least one backup copy.
• • The second check of the database involves asking for frequencies of each variable and scanning the results. This may show entries that have no business being there but which do not stand out visually, so you might not have spotted them on the first check. Examples would be an age like 16 (when all your participants were over 18), or a rating of 6 on a scale ranging from 1 to 5. To obtain frequencies in SPSS:
• • With the database on the screen, select Analyze.
• • Select Descriptives, then Frequencies.
• • Move all variables into the Variables box, then click on OK.

Scan for unlikely or impossible values, such as ages like 2 or 99 when all participants were aged 18-60, or values like 0, 6, 25 or 3.341 on a rating scale involving only the whole numbers from 1 to 5. Locate all errors on the database - this is easier to do on a paper database. You should then look at your original records, note the correct entry, and when all corrections have been made on the paper database, amend the computer database as described earlier, and save. Some sample outputs are shown in Table A5.4. You might like to look at them to see if you can spot possible errors.

The ‘answers’ to the error-spotting exercise in Table A5.4 are as follows: age has two values of 2, and one of 451. These would need to be located and corrected. Religious affiliation has one value of 12, and QI has one value of 0 and one of 11.

• The third method of checking the database will uncover any errors that cannot be detected using the first and second methods - for example entering 2 when 3 should have been entered, for a rating on a 1-5 scale. This third method is best done working in a pair. One member of the pair looks at the original data, and the other at a printout of the database. One person reads out their data, and the other calls out when they spot a mis-match. Correct all errors on the paper print-out, then correct the electronic version, save it and back it up. As stated, if you are able to carry out this type of check (and ideally, you should), then the first two checks are not strictly necessary.