20th New England Isolated Statisticians Meeting (NEISM)


November 1, 2015

Now that I am at Smith, I am technically an “isolated” statistician. It’s a funny term, because I don’t feel isolated. I have great colleagues in the Statistical and Data Sciences program (Ben Baumer, Jordan Crouser and Katherine Halverson), although I am the only person with a PhD in statistics. However, there is a great mailing list for isostats and a few local meetings, so I am happy to accept the designation if it means I get to participate in these great discussions, many of which are teaching-oriented.

Earlier this month, I got to go to my first NEISM (New England Isolated Statisticians Meeting). It was great fun, and as always I got more out of the experience by livetweeting it.

We talked about many current issues that statisticians are facing, including the importance of p-values, technology in statistics, and what skills were necessary in the context of big data (surprisingly, that conversation turned to communication, not databases). We also talked about issues more specific to teaching the introductory statistics class, including the GAISE report, which promotes multivariate analysis but doesn’t take a particular stand on technology.

A couple discussions got a bit heated.


At the meeting, we spent some time debating the importance of tables (e.g. the standard normal table) in teaching.

Interestingly, some folks in the room were insistent that tables not be used, while @rudeboybert and I were both on the side of showing the tables in conjunction with technology.

(If you are looking for a good way to integrate technology when teaching about normal distributions, the mosaic package functions xpnorm and xqnorm are fantastic. That’s where the preview image on the blog came from.)


We spent some time collating conferences we liked,

and journals,


We also discussed the importance of communication in statistics and “data science.” I mentioned the SDS program at Smith, and folks on twitter jumped in to tout the benefits of communication.

While most of the isolated statisticians teach at small colleges, we also thought about the challenges of incorporating communication in large classes, and the peer-reviewed assignments in the JHU data science specializationn were mentioned.

Group work

Another discussion that got heated, particularly on twitter, was the question of group work. We had come around to the idea that group projects were probably the best way to include communication in classes. It lets students do real analysis and present the results. Grading a few team projects is more manageable for the instructor than reading many individual papers. But, we acknowledged that there is no optimal way to select groups. One participant asked,

There were many suggestions, including


Hadley Wickham immediately piped up with a link to a paper called Effective Strategies for Cooperative Learning, which includes many of the strategies we discussed both in person and electronically. The executive summary of the paper is that you have to assign groups, you might want to think about outnumbered minorities, particularly at the beginning of a class, and you should take into account peoples’ schedules.

The paper also discusses the importance of peer ratings to assess participation, and there was a suggestion in the meeting to do this rating as part of the final.

By doing this, you are able to weed out people who really did not participate. If they can’t list their explanatory and response variables without referring to notes or teammates, they didn’t participate fully in the project.

However, there was some pushback on twitter about whether forcing students to work in groups is the right answer. As the paper says, “As we tell our students, we’re sorry if they’re unhappy about having to work in teams but the truth is that our job is not to make them happy— it is to prepare them to be professionals.” @KCombinator disagreed,

They recommended a few resources by [Paul Graham](http://www.paulgraham.com/,

My question was,

What do you think? Tell me in the comments.