CRMDA Workgroups
The CRMDA "Workgroups" are voluntary study groups in which participants discuss their research projects. The Center is always open to the formation of new Workgroups. We can help group organizers to obtain space and advertise their activities.
A Google Group for CRMDA Workgroups
A Google Group named "CRMDA-workgroups" is being used to aggregate the activities of these workgroups. At the bottom of this page is the full calendar for Workgroup sessions.
If you would like to join, request an invitation using your Google account. Go to https://groups.google.com and search for "CRMDA-workgroups". Or you can contact one of the group leaders directly and they can add you to the list.
Big Data: Analysis with Many Predictors Workgroup
Where: Watson Library Room 455
Dates: Will resume Spring semester on Friday, January 25, 2019 and will occur every other Friday at 2 PM (January 25, February 8, 22, March 8, 22, April 5, 19, May 3)
Time: 2-3 PM
Leader: Ben Sherwood, CRMDA Faculty Fellow, Assistant Professor, KU School of Business <ben.sherwood@ku.edu>
Sign-On Form: If/when you attend one of the Workgroup sessions, please register your attendance by filling out this form: bit.ly/crmdabigdata. Your participation is greatly appreciated!
A common problem in the big data era is how to deal with a large number of predictors. In some cases the number of predictors can be larger than the sample size, thus making it impossible to fit classical models, such as least squares. A popular solution is to add a penalty to the objective function. This penalty can alternatively be viewed as putting a prior on the coefficients or adding a constraint to the coefficients. Adding a penalty to the objective function makes it possible to estimate models even if the number of predictors is larger than the sample size. In addition, certain penalties can be used to simultaneously estimate coefficients and do model selection.
This workgroup will cover the following topics:
- The LASSO method.
- Using penalties with grouped variables, such as categorical variables.
- How to use penalties with imputed values.
- Can inference (p-values) be done with penalized methods.
Python Workgroup
Where: Watson Library Room 455
Dates: Will resume Spring semester on Friday, February 1, 2019 and will occur every other Friday at 1 PM (February 1, 15, March 1, 15, April 5, 19, May 3)
Time: 1-2 PM
Leader: Jonathan Lamb, CRMDA Faculty Fellow, Associate Professor, Department of English <jonathanplamb@ku.edu>
Learn advanced methods in Python!
Topics discussed in this Workgroup will include:
- Topic modeling
- Principal component analysis
- Random forests
- Predictive models
Below is the Aggregated Calendar of Workgroup Schedules
CRMDA-workgroups: a Google-Group!
The calendar shown above lists events, but the messages about group activities are aggregated in a separate place, a Google Group named "CRMDA-workgroups".