To call in the statistician after the experiment is done may be no more than asking him to perform a post-mortem examination: he may be able to say what the experiment died of. — Ronald Fisher

All models are wrong. Some models are useful. ― Chip Frank

If your experiment needs a statistician, you need a better experiment. ― Ernest Rutherford

Definition of Statistics: The science of producing unreliable facts from reliable figures. – Evan Esar

There are two kinds of statistics: the kind you look up and the kind you make up. — Rex Stout

Facts are stubborn, but statistics are more pliable. – Mark Twain

Statistics are used much like a drunk uses a lamppost: for support, not illumination. – Vin Scully

I can prove anything by statistics except the truth. – George Canning

Do not put your faith in what statistics say until you have carefully considered what they do not say. — William W. Watt

Statistics are like a bikini. What they reveal is suggestive, but what they conceal is vital. — Aaron Levenstein

If the statistics are boring, then you’ve got the wrong numbers. ― Edward R. Tufte

Thanks for sharing (In reference to the post by coastalsci on 10/2/14). It was a really interesting post. Like you mentioned, statistical analysis can often be very badly executed (wrong test, violated assumptions), or misinterpreted (effect size vs p-value, misinterpolation/extrapolation), or chosen for the wrong reasons (reject an unrealistic straw-man null hypothesis, using data to support a theory instead of trying to falsify a theory as in Popper). I think another important point is that the accelerating range of statistical options (newer tests, more complex distribution modeling) can result in fashionable, trendy tests being used before their appropriate application is fully understood. Perhaps a z test, two sample t-test, simple linear regression or ANOVA are relatively unsexy compared to quadratic discriminant analysis, latent variable models, cluster analyses and Bayesian Akaike comparisons, but sometimes the simple methods are the most appropriate. It all comes down to choosing the right test for the right question, which really means that we just need to ask the right questions, and understand exactly what we are asking. So in some senses, the question (experimental design) is at least if not more important than calculating and interpreting the test statistic. So while we’re learning normal distributions and interpreting p-values, it may be prudent to look a little into the differences among a completely randomized design (CRD), a random block design (RBD) and a latin square design (LSD), the relative strengths of different post-hoc analyses, how our choices of fixed effects and random effects influences ANCOVA output, etc. etc. At the bottom, I’ve attached links to a ton of resources for statistical analysis. They’re things that I’ve learned from, and things to think about when choosing tests and analyzing results (unfortunately none of the code written in R…boo…). Statistic Links

FYI, the “All models are wrong. Some are useful.” is actually from Statistician Extrordinare George Box from

Empirical Model-Building and Response Surfaces(1987).I would also include Stachowicz’s corollary: “All experiments are right. Some are useful.”

Aaron, Thank you for making my day and bringing me laughter – loved the quotes! I think I am going to make them into a separate document that I can easily access when sinking into doom and gloom over the 607 homework.

On a serious note, one of the reasons I pointed out the education statistics is because I think it is all to easy when surrounded by bright, highly educated, highly motivated colleagues who are all as crazy as oneself and think that spending hours, days, weeks (…) collecting mussels in New Brunswick or staring at plankton under a microscope or endlessly reading dense texts and analyzing numbers interspersed with Greek letters or endless plots and charts is all actually a fun and fulfilling endeavor.

We need to remember, we are NOT normal. I come from years spent in the computer field where my work ended up focusing on trying to find processes to translate between the normal world of everyday people using computer software in their jobs to say, run payroll or find an open bed in a hospital emergency room, and that of the software programmers and engineers who where going to design the system and write the code underneath. Now IT professionals and their ilk have been shown to be MORE THAN 2 standard deviations different than the “general public” on some social/personality tests – IT folk may be different in ways other than graduate and PhD biologists and marine scientists, but we are all together in having minds that work rather differently than most.

For that reason, it behooves us to keep trying to find ways to bridge the gap and translate what we do, why we do it, and even how we do it, to the general public. I have smart friends who shake their heads when I try to explain what I am doing in this course, much less why I can spend hours happily in discomfort on a rocky coast fascinated by algae and things lacking backbones much less sophisticated brains. Why should we care what others think or understand? Because if we want to have NSF, NOAA and other funding for research and teaching, if we want to see policies that give humanity a shot at not completely fouling our own nest, we HAVE to find ways to keep bridging the gap and make science and statistics comprehensible to our friends and neighbors.

In the meantime, I get a huge kick out of learning statistics with everyone in this class – it is a joy to be surrounded by so many bright and talented and energetic colleagues. Even if we’re not normal…