Statistical analysis is for PhDs, right? Well ...

Some researcher believe that statistical analysis is something which is so complex that people should only try to touch the subject after a 5 year PhD program. Well, like any other subject, statistics has multiple levels of knowledge and by absolutely no means is it a one or nothing type of deal like many statisticians would like us to think.

For starters, let's state the obvious: A little knowledge can be dangerous if you work alone, but in modern clinical and translational research, researchers nearly always work in interdisciplinary groups with data analysis specialists. In an environment like that a little knowledge is not only a good thing, but an essential skill to establish a successful communication with your peers. Having established that knowing some statistics is better than knowing none, where should you get started?

Two areas are essential. First, you should know in which situations a statistical test should be used. Although this might initially sound like a daunting task, the basic knowledge can be simple, allowing you to open a bilateral discussion with your data analyst. The decision algorithms on how to choose a test are largely determined by the type of variables involved (continuous, categorical, representing time, etc), their distribution (normal, count, etc), and the role each variable is representing in the research question (outcome, predictor, confounder, etc). Does it mean you should blindly follow what these algorithms tell you to do. No, you are working in an interdisciplinary team and so everything can be discussed with your data analysis specialists, but knowing some general concepts is essential for you to start a discussion and not simply accept what they say as a true and "sent from heaven".

Second, it is important to understand how results from individual statistical tests should be interpreted. For example, when a survival regression model gives you a point estimate, called hazard ratio, with 95% confidence intervals, you should be able to understand what exactly that means in the context of the tables of your paper and of your specific research question. Luckily, this interpretation is easy if you have access to a few examples from previous tables and graphics from other articles using the same statistical test. Of course, your initial interpretation should always be confirmed by other expert data analysts in your interdisciplinary group.

Knowing the indication of statistical tests and interpretation of their results is just the start. That said, these skills constitute the very basic knowledge that you can use to start conducting research while working with your peers. These basic tools are all explained in detail in the context of your individual research project during our Research Coaching Program, which you can learn more about at [Research on Research].