- Thread starter Eugenie
- Start date

medical statistician for them. I have tried to ask him about what to include, and mostly what I have found out is that medical statistics, as it is really currently done, is very high powered and complicated and far beyond an undergraduate module. So I am feeling a bit daunted. I figure survival analysis is certainly one topic. Not sure about the rest.

That would be great. Is there a way to have a private conversation on this site? It looks like it, but I am new to all this.

Also, hopeless with computers. Thank god i haven't been given a CS module to teach

Also, hopeless with computers. Thank god i haven't been given a CS module to teach

a heavy stats component. They have had linear algebra and calculus in their first year, as well as an introductory module

on probability. The module I am teaching for the second time in the fall is an introduction to statistical inference. They

also simultaneously have a module on probability theory (central limit theorem, etc). The intro to statistical inference

module covers inference techniques for situations where the predictor variable(s) are categorical, by which I mean they cannot

be put into any meaningful order. It also involves an introduction to principles of experimental design. The inference

sections are for a single categorical response variable (exact and approximate methods, including Fisher's exact test,

and chi-squared...see one of my other posts), and for a single quantitative response variable, including t-tests, ANOVA

(fixed effects only), and some non-parametric techniques (Wilcoxon tests, Kruskal-Wallis). I am trying to also introduce

them to best practice for design and reporting, such as power tests for various tests (using the Russ Lenth applets online),

initial exploratory analysis of data, reporting p-values and any ways in which the data does not precisely fit assumptions,

and post hoc tests of effect size. We are introducing R in this module as well.

The semester after this module, someone else teaches them statistical modelling, which basically covers linear and

multilinear regression and logistic regression, again using R. They also have an introduction to stochastic analysis.

So medical statistics will come after all of those modules. They will thus have quite a bit of exposure already to stats, as

well as topics in maths I haven't listed, and will have already used R for two modules.

I think they are certainly up to survival analysis. I am thinking maybe the full

MANCOVA family of techniques as a second main inference topic? Not sure about that. I know there is also quite a bit to

go over about procedure, legal aspects, etc. I really need a good reference for that, and don't have one at all. It would be

very helpful to get one.

p.s. also, if I have gotten something wrong above, especially about best practice (I am trying to consolidate the

various things I have read in hundreds of sources), please tell me that, too!

at the moment far outstrips the supply, and many universities and colleges simply are not able to hire enough statisticians to meet the need for statistics courses. Thus other members of staff are tapped to cover these courses. I think most of them (us!) try very hard

to deliver a good course, and are always looking to improve.

The good news for you is that there is a HUGE need right now for statisticians! So once you have battled through your degree

with statistics, go on and get a PhD in statistics--you can become an important resource for training the next generation!

1) If they are going to do power calculations you might show them Gpower. It is free (run by the University of Dusseldorf) and commonly used.

2) Outside academics at least, SPSS and SAS are more likely to be used than R (which is pretty much never used outside academics) You might mention it to them.

3) Regression, factor analysis, and ANOVA are commonly used methods and non-parametrics are worth mentioning (if for not other reason than they are commonly not brought up elsewhere).

4) One thing I personally found useful in my graduate statistics course was strong grounding in the assumptions of methods, how to test for their violation, and address them. If you are actually going to run statistics these are critical areas and often don't get much discussion in text.

5) If they are going to write for academic journals something like APA rules on presenting statistics is useful (or whatever the Math field uses if different; APA is geared to social science researchers).

Do you have a good source on this sort of thing? I have the sense that in many situations, such issues are still matters

of debate among statisticians, which may be why I can't always find good guidelines.

I can recommend two works you might look at. The first is by two psychologists (not statisticians) but I find it the best treatment on the practical use of data I have encountered. I will warn you, however, that some of their comments (despite being academics) set statisticians teeth on edge here. It has the advantage of presenting in each method section specific output in many of the major statistical softwares used commerically (SPSS, SAS and SYSTAT the later being less well known). If you want to use data or run methods (rather than focusing on statistical theory) it is as good as you will find and covers most of the common statistical methods. It is 'Using Multivariate Statististics (5th ED) by B Tabachnick and L Fidell 2007 chapter 4 is particularly useful for screening and cleaning data.

John Fox wrote a Sage monograph Regression Diagnostics that has a lot of stuff on the assumptions of regression, how to detect violations and (its been a while) perhaps some suggestions on dealing with them. In general there are one or more Sage monographs on virutally every methods (I have a shelf of them) and they are a decent place to start on those topics. They are very commonly used in statistic classes.

Having read what you've said, they clearly already have a good grounding in statistics, which is very nice! Therefore they are probably already beyond what I teach medical students, but I will give you a few ideas of what I teach them, and also what I teach the med stats students.

For medical students:

-Introduction to different kinds of studies - explaining the difference between cohort, case-control etc.

- Sources of variation and topics related to it, so here we introduce things like standardised mortality ratios (SMR). How to compare populations.

- Specific section on cohort studies and one on case-controls

- At around this stage we introduce medical papers to them, and get them to read a few basic ones so they can see they are starting to understand certain things like confidence intervals and p values

- A session on causality or association (including confounding)

- Randomised control trials

- Systematic reviews and meta analysis

As you can see, some of that my well be too basic for your students. But at the same time it might be useful if they are new to medical statistics as all areasof statistics have their quirks.

As for actual medical statistics, we cover a variety of topics:

- Survival analysis (which you mentioned earlier, and which is one of my favourites )

- Linear models and GLM

- Modules related to programming (in Stata/R/SAS)

- Multilevel modelling

- Bayesian inference (*shudder*)

- Methods for clinical trials, so like sample size calculations, randomisation etc

- Multivariate analysis

- Decision Modelling

These are just a selection, on that course the students are given a good grounding in medical statistics, which they then build on over time.

A few books I would reccommend (which I cant see suggested here):

Practical statistics for medical research, Doug Altman (I love this book)

Medical Statistics at a Glane, Petrie and Sabin (nice simple book)

Primer of biostatistics (again a nice simple book)

The topics your students will have done are all very useful for medical statistics, some of it might be going over some of these again briefly, to put them into a medical context. For example, my first degree is in maths and I did "reliability analysis" and it took me ages to work out this was the same as "survival analysis"!

I know this is very rough, but I am happy to go over more information about any of these modules as you narrow down what you want to do, the list is endless!