While designing your course, you will write a list of ideas for techniques and concepts that you want to teach in the course. Don't worry about whether they are good or bad ideas, and don't worry about the order. Filtering and sorting come later. You can write as many techniques as you want. The amount you'll need varies a lot from course to course but is typically between two and twenty items. Read on for tips and tricks.

Add references

If you come across good ideas for techniques or concepts to teach, make a note of where you found those ideas. Make sure to reference these if they are included in your course.

Be exhaustive

Don't be afraid to list bad ideas: it's better to have too many ideas at this stage than too few. One trick is to keep listing items until they start to get silly. 

I can't think of any ideas

  • If you know what technology you want to use, try reading the documentation for it.
  • Read through the contents pages of books on the subject. Amazon's "Look inside" feature is useful for this.
  • Wikipedia has a big list of statistical articles.

Examples

From a course on human resource analytics. This example references where an idea came from.

  • How to compare high- and low-performing groups (see Work Rules! pages 201, 343)
  • How to identify the most effective recruiting channel (how does this generalize?)
  • How to determine what's driving (attrition, high/low performance, etc.)

From a course on generalized additive models. This example distinguishes concepts from skills and techniques, and explicitly rejects ideas by specifying things to leave out.

Concepts:

  • Splines, Basis Functions
  • Model formulasAdditivity vs. Interactions
  • Smoothing
  • Penalization

Skills/Techniques:

  • Fitting a GAM to data
  • Checking that a GAM fits well
  • Visualizing a GAM
  • Making predictions
  • Selecting predictor variables (maybe too much)
  • Fitting, checking, visualizing a model with binary outcomes

Things to leave out:

  • It seems a bridge too far to go too far into the "generalized" part of GAMs in this course and discuss much about outcome distributions, at least beyond a binary example. This could be a bonus section with "This will be easier if you have learned about GLMs", but I don't know if that works within your infrastructure.
  • It also may make sense to skip most of the different types of basis functions and splines one could potentially use.
Did this answer your question?