Liezel Korf Associates has been a consulting statistician to post-graduate students for the past 20 years. If there is one thing that we have heard more than we can remember, it is: “I wish I contacted you earlier”. Much heartache and disappointment can be prevented by involving your statistician in your study right from the word go. There are a number of reasons for this.
Firstly, your research instrument should be aligned with your research questions and objectives. Questionnaires and interview schedules are often designed by simply combining a selection of questions which seem relevant to the topic. This is especially the case if the questionnaire is developed by the student, and not an existing measure. Refer to future blogs for more on questionnaire design.
When I look at a questionnaire prior to data analysis, and ask the student which questions relate to which construct, to which research question, and to which hypotheses, a blank stare is often the answer. However, by then the data has already been gathered, and, inevitably, the process of “reverse engineering” starts. Suddenly, the student has to see which research questions can be answered by the questionnaire and retrospectively change the objectives accordingly. This is often a quite traumatic process for the student, who had lofty ideals for solving the world’s problems through his / her study, and now has to settle for something much less impressive.
The second reason pertains to the statistical analysis of the study. This is something which is often glossed over in proposals. A passing reference to “descriptive and inferential statistics” is sometimes all that is provided. In other cases, these are elaborated upon by listing all possible descriptive and inferential statistics there are, for “in case” it will be needed, but exactly how it will be used is often unclear. This shotgun approach which seems to say “we’ll sort this out later” may lead to a misalignment between the proposal and the actual study, with the data not being able to answer the original research questions. The point at which this is realised, is often closer to the end of the study, when the pressure to complete it is high. Making the statistical analysis of the study less uncertain and more predictable, alleviates much of the potential stress at this point. More importantly, data can be gathered in formats and at measurement levels that will be suited to the intended analyses.
For this reason, the specific statistical analysis that will be performed to answer each research question should be specified from the outset. This will also assist the statistician to quote for the statistical analysis. I often receive a request for a quote for “statistical analysis” for a masters or PhD dissertation, without any further information being provided or available at the time. This is very difficult! The cost for performing simple analyses like descriptive statistics, cross-tabulations and correlations will be much less than that for regression, mediation, moderation or structural equation modelling !
Thirdly, and related to the second point, the statistician should also evaluate the proposed methodology of the study. Just as a poorly designed questionnaire may jeopardize the addressing of research questions, so can poorly designed methodology. While some research questions is best addressed by a qualitative methodology, others suggest that a quantitative methodology may be more appropriate. If a misalignment is identified early enough this can be addressed in time, but if the study has already been conducted, it is too late!
If you need any assistance in this regard, do not hesitate to contact Liezel Korf Associates.
0846062666
