It is not necessary to delete any questions if they already have a satisfactory level of content validity, and they also have demonstrated to have an adequate level of reliability during the pilot study.

This is because by retaining all these questions, it is still possible to replace the old items in the original questionnaire with some new ones only after conducting a fieldwork to perform an extensive assessment of both validity and reliability of these original items in the questionnaire (Bujang et. al., 2016; Jalaludin et. al., 2013). By doing so, we can avoid the possibility of inadvertently changing the true meaning of each item or construct in the original questionnaire.

This is because we can always delete any ‘problematic’ or ‘useless’ items during the fieldwork phase, when the questionnaire will be tested in a large-scale study to determine its validity.

Indeed, it is often recommended for a researcher to add more questions or items to the original questionnaire, instead of deleting or changing some of the questions (or items) from the original questionnaire in order to avoid introducing errors which may arise from making inadvertent omission of these questions (or items) found in the original questionnaire. This is because the reliability of a questionnaire will always be amenable to further improvement when it is being tested in a larger sample of respondents during the fieldwork phase, which also explains why it is always recommended for all the original items in the questionnaire to be retained for any subsequent testing and/or refinement during the fieldwork phase.

Therefore, as long as the reliability of a questionnaire has been found to be satisfactory during the pilot study, it is recommended not to delete any of the ‘problematic’ or ‘useless’ items in a questionnaire because it shall always be amenable to further refinement during the fieldwork phase, when the validity and reliability of the questionnaire will be tested by using a larger group of respondents.

Nevertheless, if a pilot study has found that the reliability of a questionnaire is too low, then it will be necessary to first revise the content of a questionnaire before subjecting it to a field-test for its validation. Therefore, the findings obtained from statistical tests for the determination of reliability of a questionnaire during the pilot study shall serve to inform a researcher whether it is necessary to revise the content of a questionnaire first, before conducting a fieldwork by using the same questionnaire in order to confirm its validity.

In summary, for all those questions that already have a satisfactory level of content validity, there are four possible scenarios as regards their internal consistency or reliability:

Scenario 1:

For those questions with Cronbach’s alpha > 0.7 and also a corrected item-total correlation > 0.3, it is recommended to retain all these questions for the fieldwork phase for validation purposes.

Scenario 2:

For those questions with a Cronbach’s alpha < 0.7 and also a corrected item-total correlation < 0.3, it is recommended to delete them because there is no relationship between scoring on those items and that on the other items of the scale (and also the level of consistency between its items in the domain is too low).

Scenario 3:

For those questions with a Cronbach’s alpha > 0.7 and also a corrected item-total correlation < 0.3, it is recommended to delete them because there is no relationship between scoring on those items and that on the other items of the scale, even though there is adequate consistency between the items in the domain.

Scenario 4:

For those questions with a Cronbach’s alpha < 0.7 and also a corrected item-total correlation > 0.3, it is recommended to rephrase some of the questions because there is insufficient consistency between some of these questions in the same domain.

However, since there is a sufficient degree of relationship between scoring on those items and that on the other items of the scale (as indicated by a corrected item-total correlation > 0.3), therefore we should still retain the questions in the questionnaire.

In conclusion, although we should retain all the questions in the questionnaire, it is necessary to rephrase some of these questions because they do not seem to have a sufficient degree of consistency between them.

In summary, the final decision to delete any questions in the questionnaire should only be made after performing appropriate statistical tests, and the results obtained from such statistical tests have provided sufficient grounds for the deletion of these questions. For example, if the corrected item-total correlation for some of the items (in the questionnaire) is less than 0.3, it is recommended to delete them because the correlation between these items’ score and the scale score is too low. This means that the distribution of scores on each of these items does not correlate with the total score. Since those items with a very low corrected item-total correlation (CITC) of less than 0.3 are those questions that are not representative of the entire set of questions, and there is no coherence between those items with a CITC of less than 0.3 and the other items in the test (i.e. questionnaire), and so they should therefore be deleted.

If you require any further clarification, please feel free to refer to the two e-books attached herewith.