Ask statistical questions

Learn how to convert your curiosities and expectations into questions that can be answered or proved right statistically with data.

Asking statistical question is the first step of the data analytic workflow. Without proper questioning, your research might get off the right track and lack of clear direction or goal. Formulating a well-defined question will let you stay focusing on each question, thus your way of gathering data and analyzing results could be much definite and clear.

From topic to statistical questions

When thinking about statistical questions, think in these steps:

Step 1: Identify your topic

Firstly, you may want to ask some question that is unknown or limitedly known in your field, or you may find some feasible intuitive thought about some interesting conclusion in your field that came up in your head last night. Be free and open in this step.

"Imagination is more important than knowledge." - This is what exactly Einstein really meant to tell you about!

Step 2: Convert that topic to a statistical question

This is the first step to answer your curiosity and proof your brilliant theory in your head. If you have your research topic in your mind, now be descriptive and scope the work. If you have that hypothesis or expected claim, convert it into a question and give it a limit to prevent overclaiming. Think of these keywords:

Population

Population is the thing of interest you want to explore about. It can be people, patients, cells, signals, modalities etc. Then, you should consider limit it to some specific group such as gender, age, diagnosis, country or area, organ, etc. Then you should be good.

Factors

Factors are characteristics of your population that you want to study about. Make sure that it is standardized and measurable.

Relationships

Relationship of the factor is the main point in most of the statistical questions. Those relationship may include but not limited to:

• Descriptive There are no relationship. You just want to collect the data from a population and report the interesting statistics about their characteristics.

• Inferential You are going to prove whether some factors may cause, relate, associate, or being a risk of other factors in those population. Some factor could also be set much alike to ensure that such effects are not caused by external "hands".

• Modelling Some of you experts may want to simulate the power of God by writing a statistical model: the equation, framework, or "black box" that predict an unknown with some known factors.

Your research may have more than one statistical question to cover whole of your hypothesis. This phenomenon reflect how large of your scope is, and how many objectives you have. We recommend you limit your scope as less as possible with your expertise, your colleague's expertise, and your limited resources such as money and time. Remember, good research focus on its quality, not size.