Understanding types of data in biomedical research: An interview with Dr Monika Soboccan
Biomedical research typically involves a number of variables; researchers often focus on exploring the relationships between them or assessing whether these variables differ between specified groups. Before running any kind of statistical analysis, it's important to determine what types of variables you have in your dataset, so that you choose the right kind of descriptive statistics, inferential statistics, and graphical representations. Today, we chat with Dr. Monika Soboccan about the different kinds of variables and how to work with them.
Dr. Monika Soboccan is a medical researcher and educator with over 5 years of experience in conducting quantitative and qualitative research. She has communicated the findings of her research in high impact-factor journals and is highly involved in providing and evaluating medical education.
1. Can you tell us about the different types of variables and examples of each in biomedical research?
You need to differentiate between different variables in biomedical research. Categorial variables are the overarching group of variables. They can be divided into nominal or ordinal variables. Nominal are the ones where there is no natural ordering (demographical data usually). Ordinal are the ones where you rank things and is a natural ordering. But you also have quantitive data, which can be discrete or continuous. With discrete data, you could talk about the number of medications a patient is taking. In continuous quantative data, you would have a range of values. Then, you also have binary variables, which are usually yes/no variables.
2. Why does type of variable matter when choosing a statistical test?
Each type of data has its own characteristics, and you need to understand the distribution of that data pool and rules of that data in order to have valid data.
3. Is it possible to convert one kind of data into another? And how?
You can regroup data by pooling them together based on their characteristics. It is possible, but then, the characteristics of the new data pool apply.
4. Supposing one set of variables in a study are of one type (e.g., categorical) and the other set of variables are of a different type (e.g., continuous). How should the researcher conduct the analysis?
For categorical values, you could, for example, assess data on count, percentages, and confidence intervals, which tend to give you a lot of additional info for your studied group.
5. Are there any special ways of handling ordinal variables? Do they have to be analyzed differently from other variables?
Ordinal data is the one that you would usually test when considering different hypothesis. Standard parametric statistical tests cannot be applied to this type of data (such as t-test and ANOVA), so this testing should be carried out using only non-parametric tests such as the Mann-Whitney U test or Wilcoxon Matched-Pairs test.
6. What are some of the common graphical methods for displaying different types of variables in biomedical research?
Based on different values, you would either decide to use scatter plots or suvival curves as part of biomedical data displays.
7. What is the difference between univariate and multivariate analysis? Could you give examples of research questions that would require each type of analysis?
The basic difference between the analysis types is the number of variables needed for the analysis. A univariable analysis looks only at one variable; in a multivariate analysis, there is more than one variable. The main purpose of a univariate analysis is to describe that varibale, while in multivariate analysis, the main purpose is to study the relationship among them. In biomed research, the purpose could be to study differences in risk factors for a certain disease.
You're looking to give wings to your academic career and publication journey. We like that!
Why don't we give you complete access! Create a free account and get unlimited access to all resources & a vibrant researcher community.
Subscribe to Conducting Research