Omega Statistics Blog

Developing a Research Question: The Testable Triad

You’ve finished your literature review. Great News! Now you can develop a research question for your dissertation or research!

If that last statement gives you a sense of dread rather than anticipation, it isn’t as bad as you think.

Think relationships. Most studies, at least the quantitative studies, have at the minimum two concepts or variables. The trick is to think of a way to test the relationship between them. The variable can be defined simply as follows:

Independent Variable: This can be a grouping variable like gender (male vs. female), or a predictor variable such as intelligence quotient (IQ) or age.

Dependent Variable: This is your outcome variable, or the endpoint you are testing. An example might be emotional intelligence. Or perhaps in a clinical study, the endpoint might be weight loss.

Two variables is a good start, especially for clinical research. For example, looking at the effects of a weight loss product (change in weight, the dependent variable) between genders (the independent variable) works great. Easy-peasy.

However, testing the relationship between two variables is usually not very compelling in a dissertation framework. But, including a third variable can make all the difference. This concept, the Testable Triad, will add sizzle and depth to your research question.

Let’s look at the process of using the Testable Triad, using some of the variables I’ve mentioned above.

I want to see if there is a relationship between gender and emotional intelligence. That is nice, but it has been done before and my literature review shows that women tend to do better with empathy, and males tend to do better with management of negative emotions. So perhaps I need to look at just one aspect of emotional intelligence as my outcome.

So I will choose empathy as my dependent variable. And now my two-variable research question is:

Two-Variable Research Question: “Do males and females differ on their levels of empathy?”

It is kind of boring, and it doesn’t contribute much to what is already out there. So, let’s consider a third variable that might affect or change the association between gender and empathy. This third variable can be thought of in one of two ways, as a mediator or as a moderator.

Mediator Variable: When included in a model, the mediator variable will account for all of the relationship between the variables of gender and empathy, or will partially account for the relationship

Examples of mediator variables could be IQ, age, or perhaps age group. For instance, levels of empathy may have nothing to do with gender once you take into account an individual’s IQ level or age. I decide to use IQ as a mediator in my testable triad. And I would word my research question as follows:

Testable Triad Research Question: “Does IQ level mediate the relationship between gender and empathy?”

In plain English, I would be testing to see if IQ totally, or partially, accounts for the levels of empathy. It could be that gender doesn’t matter at all once you take into account IQ. Now that is much more interesting, isn’t it?

Moderator Variable: A moderator variable can affect the magnitude or even the direction of the relationship between two variables. Often, a moderator is a grouping type of variable, such as age group rather than age. But not always!

Examples of moderator variables could be: IQ classification with 4 levels (below average, average, above average, exceptionally high) or maybe age group with two levels (below 40 years of age, 40 year of age of older).

As an example, the relationship between gender and empathy may change according to the level of IQ of a person, or according to the age group a person belongs to. And I could word my question as follows:

Testable Triad Research Question: “Does IQ classification moderate the relationship between gender and empathy?”

In plain English, I would be testing to see if the difference in empathy between men and women is moderated (changed) according to what IQ group the gender groups were in. Maybe those with lower IQ levels have more empathy no matter what gender they are, but people with higher IQ levels are not very empathetic. OR maybe the women stay empathetic but the men change depending on IQ level. OR… well you get it, right? IQ level may be doing some interesting things to that gender/empathy relationship.


    Think about YOUR study. Depending on what your literature review revealed, the theoretical framework of your model, and the gap(s) in the research, you could think of many “third” variables for your testable-triad.

    Make a list of these possible mediators and moderators, and choose one (or more, your research question doesn’t have to use only three variables, it doesn’t have to be a triad…but don’t overdo it, remember to keep it simple yet informative) that can be used to tweak what is already in the literature base a bit to look at a concepts in a more in-depth or slightly novel way.

    A final note: The clinical research example I presented earlier, with gender and change in weight as variables, could also benefit from the testable triad: I could use gender as the independent variable, change in weight as the dependent variable, and hours of weekly exercise as a mediating variable…or perhaps exercise type as a moderator.

    I hope the testable triad proves useful to your research.

  • Want to learn more and keep in touch? Sign up for the Omega Statistics email list. You will receive a quarterly newsletter and notifications of events such as our FREE Stats for the Masses webinars!
  • Please visit the Omega Statistics Events page to access previously recorded webinars, and to sign up for the upcoming events.
  • And, you can always send Elaine Eisenbeisz an email at or give her a call at 877-461-7226 during business hours (8 a.m. – 5 p.m. Pacific Time, Monday-Friday). She’d love to hear about your research!

Bridging the Gap: Choosing a Relevant Dissertation Topic

Finding a suitable topic is the first of many challenges you will face in the dissertation process. Where do you even begin?

The best place to start is with your own interests and the questions that have arisen for you over the course of your studies. You want to have familiarity with your subject matter. You will be working with it for quite a while!

But what about that elusive gap in the literature? There are two valuable resources, (1) your committee and (2) the literature.

Your Committee, Gotta Love ‘Em!

Often a committee member or two will be actively involved in their own research and may be open to offering suggestions for further pursuing their research via your dissertation work.

Although working on a committee member’s pet project sounds like an easy option, and it is in the beginning, it can become cumbersome for you as a researcher. A committee member with a vested interest in your success to enhance his/her research goals may be compelled to direct (meaning: control) your research more than you’d like. And often you will end up feeling like a research assistant who is only working to further the committee member’s goals and agenda.

Having said this, many researchers do take on a committee member’s research and survive the process. But don’t say I didn’t warn you!

The Literature, Your New Best Friend

I cannot stress enough the need for a thorough literature review in your area of interest. Even if you have an idea of your topic, a literature review is necessary to:

  • Confirm interest in your planned topic, not just for yourself. Do others care? Is it considered important, even if by just a handful of people? Who are these people?
  • A literature review may also show you that work on your chosen topic has been exhausted and that the gap you were considering has been filled. No worries, it is better you know now than after you’ve written and submitted your proposal.
  • Inform you of the most recent findings and gaps in the research. The gaps can be found by reviewing the “implications for further study” in the Discussion sections of journal articles and other dissertations related to your planned topic. You could be just the person to address that need of further study, and in deciding to do so, BOOM, you have found your problem statement and literature gap!
  • A literature review is a possible first step in establishing relationships with renowned and current researchers in the field.
  • Why re-invent the wheel? Think about replicating a study, or gently tweaking a study, that you’ve found in your literature search. It is not plagiarism if you ask permission from the author(s) of the original study and if you note that you are replicating the study, with proper citations, in your work. Check the “suggestions for further research” sections of discussion chapters to see what the author thought were the next steps research. One of these “next steps” could be the basis of your study! For the most part, researchers love it when someone wants to replicate their work or use their instrumentation in a new study!

Some General Advice

  • I mentioned earlier that you should be familiar with your topic because you will be working with it for quite a while. Also, your topic should be something that you are interested in. The time and money involved in the dissertation process is like remodeling a kitchen. It takes twice as long and costs twice as much as you think it will. So a topic that interests you will be much less mind numbing for you.
  • However, don’t study something that makes you emotionally upset (or joyful) or for which you have already formed a strong opinion. Bias is not a good thing in research and this is probably the best way to introduce it into a study. So avoid topics that are emotionally charged for you, no matter if the charge is positive or negative.
  • Parsimony is our friend. Only one gap needs to be addressed. So don’t attempt to fill in all the empty spaces. A big saying in my field is, “The best dissertation is a done dissertation.” Some of the best studies are simple and specific. And done!! Here is where the skills of a statistician who knows how to design a solid and robust research are needed. So keep my email and number handy and call when you are ready to design your study.
  • Don’t design your own survey instrumentation unless you absolutely on your life have too. There are all sorts of survey tools for any number of topics. So try to avoid designing your own unless you want to run a validation study for your tool inside of your actual study. Instead, search the literature for some tools that might work for you, and ask for permission from the designer of a tool to use it. The author’s email and place of work is usually right at the bottom of the first page of the journal article. Reach out. Most researchers love to see their instrumentation re-purposed because it adds to the validity of their tool.

Next Steps?

    The second major challenge in the dissertation process is taking the research topic and designing a workable study around it. In a future post I will share some tips on how to turn your research topic into a research question.

    I am also planning a webinar to give you some tips and resources for refining your topic into a workable study! I’m still working on it, so keep in touch!

    In the meantime, check out a past webinar, Overview of Research Design, in the archives of the Omega Statistics channel on YouTube. And remember to sign up for the FREE literature review webinar in September, 2017. I look forward to seeing you there!


A man jumping over a ledge with the word Success on it

Instead of writing about a statistical concept in this post I wanted to share something I’ve discovered about myself and other successful people I’ve worked with. I am hoping you will also find it useful and fun…

I find a large part of being a private practice statistical consultant is talking people down from the…

Continue Reading


Zeros and Ones in a pile.

Coding groups on nominal (categorical) groups of an independent variable is not a problem in analyses such as t-tests or ANOVA. However, things can fall apart fairly quickly if nominal variables are not coded properly for use as predictors in linear models such as multiple regressions.

Let’s take a simple example of a variable for…

Continue Reading


Students Working On Code

One interesting and very useful aspect of dummy coding is the ability to use these dichotomously coded variables as predictors in a regression model.

A regression model does just that, it models something, represents something.

Let’s investigate a hypothetical model of stress.  We want to see if biofeedback training and gender have an effect on…

Continue Reading



Thank You - Watch Your Inbox For Our Next News Letter

Accreditations and Memberships

Murrieta Chamber Of Commerce Logo
Better Business Bureau Logo
American Statistical Association