Posterior samples

Open Access publishing is an exciting new development but like any new industry there are those who would seek to unethically make a quick fortune by providing a sub-par service. Science did some research into this by submitting a fake article with a number of glaring errors (such as non-existent institutions) in it to a number of journals around the world to see who’d accept it. The results are quite interesting. I know of a few instances of work published in a respectable journal being plagiarised and published in one of these scam journals and one case where the authors hastily retracted their submission once they found out the journal wasn’t anywhere near as prestigious as the editors made out. I support the goals of Open Access and web-based publication but be careful.

I’ve been poking around the QUT Wiki looking for information on graduation. In the Higher Degree Research portal there’s a link to a brilliant article (three years old) on how not to write a thesis. Probably good advice for any PhD student looking to graduate sooner rather than later.

Elizabeth C. Matsui has been working with Roger Peng for several years; she has some advice on how to cultivate a successful relationship with the biostatistician you’ve brought on to your medical science project. I think some of these could be adapted to be more general to cover dealing with statisticians. One of the biggest things I could add here is that you need to recognise that asking someone to “help with the statistics” means dedicating some serious time and effort to figuring out a conceptual model (which will be converted into a quantitative model) that addresses the scientific questions you wish to ask. Asking a statistician to “calculate the correlations” or even more broadly “do the statistics” is like asking a scientist to just “run an experiment” or “do the science”. These things all take time, energy and communication.

Further to this, don’t just ask for p-values if you’re dealing with a statistician. P-values represent some probabilistic statement, such as “What is the probability of seeing a test statistic at least as extreme as this if the true value of the parameter was zero?”. Sure, you can do simple tests like ANOVA or Chi-Squared goodness of fit but the real value in working with a statistician is being able to develop models that represent assumptions about the data and then assessing whether those assumptions are justified and looking at the inferences that they provide. As my unit co-ordinator for SEB113 pointed out to me the other day, default hypothesis tests of H0: β = 0 at a 5% significance level may not always be sensible and may not even answer the question you’re interested in answering.

And finally here’s a video that’s almost a year old that shows a very interesting case of eliciting a distribution from the American public about what they perceive the distribution of wealth in their country is and what they think it is. Turns out the ideal is as far away from the perception as the perception is from the reality.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s