A flurry of flavour and a pimpin' jive.

If you do nothing else by the end of this week, go to asda and buy a box of frozen summer fruits. When they are thawed on the outside but still frozen in the center, take a handful and hold them in your mouth. Gradually work your way through the utterly mouth-watering fleshy parts until you are left with little rolls of flavoured ice. Even if you're allergic to all berries, do this. There are few better ways to die.

I'm in a very cheerful mood this afternoon. All my ranting about psychology sort of fell apart when my friend (who has been given an offer of PhD from everywhere, ever) asked me to help him with statistics. My process for going through these problems seems pretty nonsensical to anyone incapable of mind-reading, but I adore working out how to do it right, and then understanding why what I did was correct. I took some mad problem about normal and mutated strains of a certain string of DNA, and got a lovely precise answer. I'm sure the results might mean something to crazy scientists who understand the myriad acronyms involved.

So I have a passion for statistics. That's actually enough for me. That's a damn good skill in many areas of life. Not only because I can, you know... statisc things; but because the whole process of complex problem solving has ingrained itself into me. With that said, someone needs to explain to me the concept of statistical effect size.

I understand that effect size is a measure of the magnitude of the effect of the independent variable (e.g. caffeine) on the dependent variable (e.g. pimpin' jive). A large effect size means that caffeine really makes pimpin' jive better or worse. A small effect size means it only effects it a little bit. Fine. Good. But what constitutes a 'large' effect size. When I asked my tutors this, the question seemed entirely foreign in nature. They tried to explain it, but I was convinced they were answering "what is effect size" rather than "What is the numerical limit of effect size?"

Say that the effect size of a statistical test comes out as 0.8. That's a large effect size, and any researcher would be delighted to achieve that if it was also significant at the o.o5 level. It is rare but possible to achieve an effect size much larger than that. In various made-up tests we've examined in class, the effect size has been in excess of 10. As far as I'm aware, the size is theoretically limitless. How then, can you distinguish between a large and a small effect? If small - medium is 0< - 0.6 and large is everything above that, doesn't that seem a fairly inaccurate measure. Logically, can one independent variable not only account for 100% of variance in a dependent variable; so by making effect size 1 the maximum possible?

If anyone has an explanation I'm missing, please let me know :).

In other news - Super trendy beach shorts should have pockets, and one of the best nights I've had in a while came from a simple walk home with tremendous company.