Interpreting regression coefficients can be tricky. Especially when there are interactions in the model. Or categorical predictors. (Or worse – both.) But there is a secret weapon that can help you make sense of your regression results: marginal means. They’re not the same as descriptive stats. They aren’t usually included by default in our output. And they sometimes go by the name LS or Least-Square means. And they’re your new best friend. So what are these mysterious, helpful creatures? What do they tell us, really? And how can we use them? 

Question: Can you talk more about categorical and repeated Time? If I have 5 waves at ages 0, 1  year, 3 years, 5 years, and 9 years, would that be categorical or repeated? Does mixed account for different spacing in time?   Mixed models can account for different spacing in time and you’re right, it […]

As mixed models are becoming more widespread, there is a lot of confusion about when to use these more flexible but complicated models and when to use the much simpler and easier-to-understand repeated measures ANOVA. One thing that makes the decision harder is sometimes the results are exactly the same from the two models and sometimes the results are vastly different. In many ways, repeated measures ANOVA is antiquated -- it's never better or more accurate than mixed models. That said, it's a lot simpler. As a general rule, you should use the simplest analysis that gives accurate results and answers the research question. I almost never use repeated measures ANOVA in practice, because it's rare to find an analysis where the flexibility of mixed models isn't an advantage. But they do exist. Here are some guidelines on similarities and differences:

In a previous post we discussed using marginal means to explain an interaction to a non-statistical audience. The output from a linear regression model can be a bit confusing. This is the model that was shown. In this model, BMI is the outcome variable and there are three predictors:

We often have a continuous predictor in a model that we believe has non-constant relationship with the dependent variable along the continuum of the predictor’s range. But how can we be certain? What is the best way to measure this?

Ah, logarithms. They were frustrating enough back in high school. (If you even got that far in high school math.) And they haven’t improved with age, now that you can barely remember what you learned in high school. And yet… they show up so often in data analysis. If you don't quite remember what they are and how they work, they can make the statistical methods that use them seem that much more obtuse. So we're going to take away that fog of confusion about exponents and logs and how they work.

SPSS and Stata use different default categories for the reference category when dummy coding. This directly affects the way to interpret the regression coefficients, especially if there is an interaction in the model.

Structural Equation Modelling (SEM) increasingly is a ‘must’ for researchers in the social sciences and business analytics. However, the issue of how consistent the theoretical model is with the data, known as model fit, is by no means agreed upon: There is an abundance of fit indices available – and wide disparity in agreement on which indices to report and what the cut-offs for various indices actually are.

Have you ever experienced befuddlement when you dust off a data analysis that you ran six months ago? Ever gritted your teeth when your collaborator invalidates all your hard work by telling you that the data set you were working on had "a few minor changes"? Or panicked when someone running a big meta-analysis asks you to share your data? If any of these experiences rings true to you, then you need to adopt the philosophy of reproducible research. Reproducible research refers to methods and tools developed by large software development teams but which can help you keep a sense of order in your data, analysis programs, and results.

Think of reliability as consistency or repeatability in measurements. Not only do you want your measurements to be accurate (i.e., valid), you want to get the same answer every time you use an instrument to measure a variable. That instrument could be a scale, test, diagnostic tool – obviously, reliability applies to a wide range of devices and situations. So, why do we care? Why make such a big deal about reliability?