What should a work supervisor put in a recommendation letter for someone applying to statistics PhD programs?

Question

This fall, I'll be applying to PhD programs in statistics in the U.S. I'm currently a full-time student in an applied math master's program in the U.S.; until this fall, I was a part-time student and a full-time data analyst at a financial services company. It looks like I'll only be able to get two good recommendation letters from professors, so I'm considering asking my former work supervisor to write one for me. She was my supervisor for two and a half years, during which time we worked together closely on many projects. She doesn't have a PhD in statistics or any other field, but she was a data analyst before becoming a manager. What sorts of things should I ask her to write about in her letter? The professors I'm going to get letters from taught statistics classes that I took and are supervising statistical research that I'm doing; I'm guessing that I should ask my former supervisor to write about things that my professors might not touch on, but I'm not sure what those things should be.

Answer 1

Often, PhD applicants come along with transcripts showing solid grades in their undergraduate courses. However, this does not always translate into an ability to apply the material in real-world settings, or to find and successfully pursue interesting/important questions. These are naturally important skills for success at research, and may be areas where your supervisor's reference can be very helpful. Some specific questions that I would value seeing addressed in such a reference (which obviously may need some refinement depending on exactly what your current job involves):

• Can you work effectively without close day-to-day supervision?
• Can you come up with useful ideas/questions/tasks for yourself? Are you an effective problem-solver?
• Are you able to weigh up the pros and cons of different strategies and arrive at a sensible recommendation for what to do next?
• How do you respond when someone disagrees with you about the correct approach to something?
• Are your solutions realistic, effective and properly-tested?
• Are you able to adapt your thinking/approach as circumstances and requirements evolve?
• How do you respond if it turns out you're wrong about something or have made a mistake?
• What do you do if you're stuck on something?
• Can you explain your work clearly and concisely? To non-specialists? To experts? In conversation? In a formal presentation? In writing?
• How do you deal with colleagues? Are you collaborative? Are you good at teaching someone something new?
• Are your technical skills up to scratch?
• How do you respond when under pressure (e.g. deadlines)?

Finally, remember that examples are worth far more than unsubstantiated statements.

Answer 2

The readers of such letters will normally be looking for predictions of success in the academic program and thereafter. The past is less important (what you have already done) unless it supports that prediction of success.

If the letter writer understands that, and they are supportive, then you should be fine. But a "bald" statement that "I predict the success of this person." is too easy to discount. It needs to be a bit more subtle than that.

Answer 3

Having written letters for students outside my discipline, here are some things she might be able to write about that your professors might not.

• Ability to work collaboratively in a team.
• Ability to explain statistical concepts to non-statistician clients/team members.
• Flexibility.
• Ability to take on new challenges, such as with a particularly difficult set of data or complicated sample.
• Ability to communicate results effectively in writing and presentations.
• Organization and reproducibility of your work.
• Work with diverse groups.
• Desire and willingness to learn new things.

These are all things that a faculty member looking to bring someone into their work group might want to know and would value. If the letter writer can provide examples of this that would be very good.