Colleagues -
What are your thoughts on including statistical formulae in dissertations? Thanks!
Ruth
What do you want to know, specifically? Obviously the appropriateness of formulae depends on the relevance to the topic. If the dissertation is about statistics in particular, and/or if it's introducing a new method or using a method that is not well-known, then it makes sense to include formulae, but it's not so necessary for methods that are already well-known.
Hi, Stephen,
An unhappy student asked this question on an online blog in the last few days. Her specialization is in Leadership and her statistical model is a logistic regression. Her supervisor is requiring her to state the statistical formulae, and has provided the write-up for her - which she does not understand. While she was introduced to various statistical models during her PhD coursework, they did not get into the statistical formulae. She has examined quite a large volume of dissertations and have not seen the use of these formulae in the dissertations. Since she does not understand what they mean, she does not want to include them in her dissertation. I was curious about this as I did not have this in my dissertation, nor did my peers.
Ruth
If she is using a statistical method in her dissertation she should probably understand the formulae behind it. So rather than worrying over whether or not it's necessary to include, my recommendation to her would be to just include it (at least in an appendix or something; if the committee asks for it then it's easiest to just do it) and also read up on logistic regression, which is the important thing anyway. (The chapter from Tabachnik & Fiddell is a good resource, that's where I learned logistic regression.)
To be fair, this depends to some extent on just what formulae we're talking about. For example, I use logistic regression (or its mixed-effect equivalent) quite a bit, and I know the logit and inverse logit functions (they're necessary to understanding the output of the model anyway), but I confess I don't know the details of the various implementations of maximum likelihood estimation. Like, if you asked me to write a program to do logistic regression from scratch, I wouldn't be able to (although I'm sure I could find some simple resources and figure it out in a day or two), whereas something like ANOVA or OLS regression I could. I feel that knowing the conceptual formulae is necessary but knowing the implementational details maybe not. But that might just be me, though.
Thank you, George and Stephen.
I understand all points that you made: sometimes it is better to just do as the supervisor says, I suppose. From what I understand, she was taught a basic overview of the different types of statistical methods, but that did not include delving into the formulae to the extent that she could relate her analysis in those terms. Thank you for both of your perspectives...
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics