10.4 Modifying the script once interviewing has begun

This is an onerous situation best avoided at all costs. Really. Just make sure you never have to do it. Ever. But you've done it anyway. Oh well. Here's how to fix things.

Ideally a survey should be preceded by piloting and dress rehearsals. If an error is still present but is only identified when the survey has started then it may be necessary to alter the script after data has been collected. Obviously, if the survey data collected are not considered important and may be erased, then the script may be modified and recompiled without difficulty in the usual manner. Otherwise, proceed as described below.

If the set up has been done as suggested in this manual then the current survey data will be on the share server hard disk, drive S:, and the project's development area will be located on the network, drive X:. Therefore, the compiler can still be used to process the files in the project's development area without risk to the data sets on the share server. This allows the script to be modified without risking the current data. Once the script has been corrected it is necessary to copy the share server data to the project's development area and recompile without erasing the data.

Recompiling Procedure:

The process is summarised below.

1. As stated above, all changes to be script should have been made and tested within the project's development area (X:).
2. Stop all interviewing.
3. Back-up all survey data.
4. Copy the survey data sets ANSWERS, CONTROL, PROTECT, CONFID, and CALLS from S: to X:.
5. Run GEIS without using the share server. You should be accessing X:.
6. Import the new script.
7. Open the compiler window.
8. Clear the check box labelled Erase survey data.
9. Compile the script. The actions taken by the compiler are listed below.
   1. The old data sets are saved to temporary data sets.
   2. Compilation proceeds as normal.
   3. The old data sets are used to update the new data sets.
   4. A comparison is displayed showing the old and new data sets.
   5. Unlike a normal compile the new data sets are sorted by ID.
10. If the compile fails the old data sets will be restored.
11. After a successful compilation the new data sets may be copied back to the share server (X: to S:).
12. Resume interviewing.

Dealing with partly-completed interviews after re-compiling:

The modified script logic may be incompatible with the old script logic structure. For example, a new question may have been introduced, omitted, or re-ordered, the logic structure may be changed, and so on. In such circumstances it is not possible for the compiler to predict where interviews that had been partly-complete when the change was made should be resumed after re-compiling. Therefore, the compiler resets all partly-completed interviews to resume from the beginning of the script. The correct point at which a partly-completed interview should be resumed must then be determined manually.

To do this:

1. Manually restart each partly-completed interview.
2. Press the Next button until no further progress can be made. The following message is displayed: ``Unable to jump. Check item and prior items answered.''.
3. Exit the interview.
4. Save the interview status as PQ.

General comments on re-compiling:

Using a modified re-compiled script can have numerous repercussions. Since there are many changes that could be made to a script only some general advice can be given below.

• Changing the data type of an item in a script may affect how the data is stored. If changed, the previous data will not be retained by the old item.

• Changing the number of options, option codes, or option texts in CHCE or MULT items does not produce an error. GEIS will report when code values in the old versions of CHCE or MULT items are not found in the new script versions. Depending on the changes made the new data collected may not be comparable to the old data.

• Changing an item's name will prevent the old data being imported to the correct item.

• If the names of two items are interchanged, their data may be swapped.

• New script items will result in additional variables being created in the ANSWERS data set. These variables will be blank for interviews that were conducted before the script was changed.

• The number of question lines or question text may be changed without any difficulty, but the data collected following the change may not be comparable to the data collected afterwards.

• Changing an item's informat may result in different types of data being entered.

• Changing an item's format, label, module, submodule, reasonableness limits, absolute limits, or label will probably not present a problem.

• Changing an item's SPSs may result in considerable changes in the script's logic structure.

• It is better to thoroughly test a script beforehand than to make changes later!