“I am Doing Research, What is the Most Important Thing?”

I often get this question, directly or indirectly. Directly, when people come to me and utter those exact words. Indirectly, when they ask a slightly different question, which actually denotes a lack of clarity of what the real question should be. My typical answer is: what is your objective?

It is very common for a researcher or a postgraduate student – who is compelled to conduct a study – to ask things like, ‘Which statistical test do I use?’, or ‘What is the most suitable study design?’, or ‘How many respondents do I need?’ No doubt, these are important and valid questions. But the answer to all of them is the same: know your research objectives. In other words, that is the most important thing when doing research – know your objectives very clearly, and specifically.

Being clear with what you want – or spelling out your study objectives – is the first step that helps you in formulating subsequent ones. Name it – calculation of sample size, choosing the study design, deciding what tool to use, determining the right statistical approach – they all depend on what you want to achieve in the study. For instance, you plan to measure the prevalence of (low) health literacy among older adults who visit the geriatric clinic in your hospital. What is the best study design? Since you are measuring prevalence, perhaps it’s cross-sectional. What tool do you choose? Since you want to measure health literacy, you need to browse through the different possible questionnaires meant for health literacy and find the most appropriate one, especially for your targeted age group. What statistical approach should you go for? If the objective is to measure prevalence and nothing else, you may limit your scope to purely descriptive statistics.

Let me share a second example. Now your goal is to see if there is a relationship between getting cancer and falling into depression. Let’s rephrase that a bit to make it clearer – you want to see if getting cancer causes depression. Here you need to be specific, because how your objective is phrased determines the subsequent steps, as mentioned above. Are you trying to ascertain if cancer causes depression, or are you trying to find an association between cancer and depression? They both differ – in the former, you want to know if people get depression because of cancer. In the latter, you want to know if having cancer is associated with depression. The word association does not amount to causality; it may be suggestive, or it may sometimes indicate something different. It’s like saying someone who loves coffee usually wears black. Does drinking coffee make the person wear black? Not really. But you observe that coffee lovers tend to wear black more often than others. It’s not causality. It’s association.

analytics text
Photo by Timur Saglambilek on Pexels.com

Let’s consider both scenarios. First, you want to determine if cancer causes depression. As I said, that statement pretty much has given you many clues to your next steps. How do you calculate the sample size? You need to know the OR (odds ratio) or RR (relative risk) from previous studies, before you can do the maths. What study design should you opt for? Since you are talking about ‘causality’, the best might be a cohort or longitudinal design. What tool do you use? Look at your objectives – you mention ‘cancer’ and ‘depression’ so you need to think of ways to measure that. It can range from a simple method like a questionnaire to something more complex like medical records or clinical diagnosis. The approach is rather different if you decide to investigate the association between cancer and depression. Association generally means a cross-sectional study would suffice. The tools used might be the same as previously mentioned, but be cautious when interpreting your results. As for statistical analyses, you can consider bivariate analyses and regressions for cross-sectional design, and more advanced techniques for longitudinal data such as regressions or generalized estimating equations. In addition, for cross-sectional studies you may have to worry about response rates, while for cohort studies your main concern is usually attrition.

In conclusion, the most important thing to know and be aware of, when conducting a study is your objectives. Be very clear and specific about them. Make sure that your study design, sample size calculation, statistical approach and data analyses will eventually be able to answer your research questions. Whenever in doubt, ask yourself: what do I want from this study?

Raudah Mohd Yunus


Share on facebook
Share on twitter
Share on linkedin
Share on whatsapp
Share on telegram

Bincang dan komen

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.