Count Models – Enrollment
Have you ever worked with a dependent variable that is a discrete count of some quantity?
Number of days, number of symptoms, number of incidents, number of arrests — count variables show up across many fields.
They’re truly numerical, so parameters like the mean make sense, but they’re not continuous. They can look normally distributed, but they aren’t.
They have limitations that normally distributed variables don’t:
- They can’t have negative values.
- They can only have values with whole numbers.
- They often have a mode at the end of the data set (usually at 0).
There is a whole family of models made just for count variables. These models can account for the limitations, common distributions, and unique issues that arise with count variables.
This workshop will guide you through them.
You’ll learn why linear regression doesn’t work well (even with a transformation), and what does. You’ll learn 6 different types of count models (plus all their varieties) and how to decide which one works best in your specific data set.
After taking this workshop, you’ll be able to recognize the need for, conduct, evaluate, choose, and interpret a count model — and use it to predict outcome probabilities.
- Begins: January 12
- 12 hours of live instruction, plus 6 Q&As
- 1-year access to workshop website
- Instructor: Jeff Meyer
- Stat Software: Demonstrations in Stata, SAS, and SPSS
- Level: Advanced (Solid experience running linear models is required)
- Investment: $497 / $297 (Student)
About the Instructor
Jeff Meyer is a statistical consultant, instructor, and writer for The Analysis Factor.
As your workshop instructor, Jeff will be teaching the workshop via live webinar sessions. He’ll be writing the code for Stata to use in the examples and exercises. He’ll also be at the live Q&A sessions to answer any questions you have.
For Jeff, being an effective instructor is about more than just knowing the subject matter and applying a logical approach to analyzing data: He also enjoys working with people and, most importantly, he cares about your success.
He has an MBA from the Thunderbird School of Global Management at Arizona State University and an MPA with a focus on policy from the Wagner School of Public Service at New York University.
About the Software Specialist
Audrey Schnell is a statistical consultant at The Analysis Factor. She has particular expertise in biostatistics, including Inter Rater Reliability, Case control studies, and linear models.
As your software specialist for this workshop, Audrey will be creating the SAS demonstrations and code for examples and exercises. She’ll be at the live Q&A sessions to answer your SAS questions.
Audrey has a Master’s Degree in Clinical Psychology and a PhD in Epidemiology and Biostatistics.
Comments from Past Participants in Jeff’s Workshops:
6 Live, Interactive Webinar Sessions with Jeff Meyer
Sessions start at 12 pm (US EST) and last 1.5 – 2 hours, depending on the number of questions.
6 Live Q&A Sessions with Jeff Meyer and Karen Grace-Martin
Q&As are from 12 – 1 pm (US EST).
Exclusive Access to a Participants-Only Website, Your Home Base for the Workshop
You’ll find everything you need here:
Each topic we cover will include demonstrations in Stata, SAS, and SPSS.
✅ Understanding Count Models
We’ll start with an understanding of what counts are and why they don’t work well in linear models. We’ll talk in detail about why a count model is necessary and what a log link means.
Then we’ll go through a brief overview of the most important concepts and steps so you have a big-picture understanding. This lays a strong foundation for the rest of the workshop.
- Understanding what counts are
- Why OLS linear regression doesn’t work
- Model distribution curves
- Model assumptions
- The modeling process
- Maximum likelihood estimation
- Generalized linear model (IRLS algorithm)
- The basic Poisson model and the log link function
- Exposure variables: rates vs. counts
- Understanding and testing for overdispersion
✅ Understanding the Various Types of Count Data
The tricky thing with distributions of count data is they’re not all the same. Some have more zeros than we’d expect; others none. As we explore these patterns, you’ll learn how to recognize which one you’ve got, like:
- Excessive zeros (zero inflated)
- Excluded observations (truncated)
- Aggregate observations (censored)
- Bounded and unbounded (hurdle)
✅ Reviewing Specific Count Models
As we go through the workshop, we’ll discuss in detail quite a few distinct models. For each one, you’ll learn when a model is appropriate (and how to tell), how to interpret output, and how to build and assess a well-fitting model that tests the hypotheses you’re interested in. These model types include:
- Negative Binomial
- Zero inflation
✅ Understanding the Model Output
There are various ways to investigate in detail the effects of individual predictors: graphing, calculating predicted counts and examining rate ratios.
We’ll spend a lot of time on interpreting coefficients, especially for interactions, which get even trickier. We’ll cover:
- Interpreting coefficients and rate ratios (details including interpreting interactions)
- Marginal effects for predicted counts
- Partial effects
✅ Evaluating Model Fit
Now that we’ve got a solid understanding of each model, we’ll learn ways of evaluating it.
Is it any good? Does it fit the data? How well does the model predict outcomes? What happened to R-squared?
You’ll learn how to evaluate the model with:
- Likelihood ratio tests
- Information criteria: AIC, BIC
- Analysis of residuals
Is This Workshop Right for Me?
This workshop is for you if:
- You have tried to do a Poisson or negative binomial regression before, but found it confusing or difficult, and don’t really understand it.
- You have used linear regression or ANOVA, and you want to expand your knowledge.
- You know you will need to implement a count model soon.
- You want to expand your statistical capacity.
We’ll do demonstrations in Stata, SAS, and SPSS. It’s ideal if you know one of those, but the focus of the workshop is on the meaning, steps, and interpretation of count models regression models, not on the software.
So if you’re familiar with another software package and are comfortable reading the manual to understand the defaults in its logistic regression procedure, you’ll be fine.
A special note for SPSS users: While SPSS will do most of the models we cover, it can’t do all of them. So at some point you may need to use another package.
This workshop is not for you if:
- You are a statistics beginner and have never done a linear regression. You should be familiar with interpreting regression coefficients, p-values, R-squared, dummy coding, and interactions.
- You have an advanced degree in statistics and are looking for a theoretical course on the topic. You’ll be disappointed in the lack of proofs, calculus, and equations.
This is an advanced level workshop.
But it’s designed for researchers, not PhD statisticians.
You should have solid experience running linear models. Experience with ANOVA or linear regression is fine — as long as you have a basic understanding of least-squares estimation, interpreting coefficients, dummy variables, and interactions.
Familiarity with logistic regression will be helpful, but not necessary. Many of the concepts you’ll learn here are similar (though not all identical) to those in logistic regression.
Plan to set aside 5-8 hours each week.
Aim for full participation in webinar sessions, Q&As, and exercises.
You’ll learn the basic concepts and get some clarity even if you don’t do the exercises and run the examples on your own.
But you won’t entirely get it unless you get your hands dirty with some data. That’s what this workshop is for.
Sorry, this workshop is full, and registration has closed.To get first notice when it opens again, please join our Advance Discount List. Check out some of our other online workshops too!
Your registration fee is fully refundable up to 72 hours in advance. Because enrollment is limited, no refunds will be granted after the program begins.
As with all our programs, your satisfaction is guaranteed. If you participate fully in this workshop — watch the videos, read the materials, complete the homework, and try out everything included — and find you are not satisfied for any reason, we will give you a full refund, no questions asked. Just notify us within 90 days of purchasing the workshop.