GamerGate’s Spectacular Display of Numerical-Based Tom-F*ckery
Hey there, fellow literates! Thank you for taking the time to read this, my first Medium article! A few months ago, Newsweek wrote an article hypothesizing GamerGate is more focused on harassing females than championing ethics in gaming journalism. They used a questionable methodology; so making a competent counter-argument should have been simple for an average statistician. Fortunately for us @Cainejw, who I’ll refer to as JW from this point forward, stepped up to defend GamerGate’s honor. It turned out JW is as inept at doing statistical analysis as Gamergate is at convincing the general public it’s really about ethics in gaming journalism. JW wrote a statistical analysis rebutting Newsweek’s article, and the rest of my article will be dedicated to ripping JW’s analysis to shreds. I’m somewhat late to the party, as @Zennistrad has already called out JW on his statistical shenanigans months ago. They astutely point out the failings of both JW’s per-capita nonsense and his sentiment-analysis-for-dummies methodology. Best of all, they prompted JW into writing an appendix to his original rebuttal that can best be described as a spectacular display of pure, unadulterated, numerical-based tom-fuckery.
Unlike JW, I can actually math.
Before I get ahead of myself, let me clarify this article is not meant to show that Newsweek’s GamerGate article is rock-solid. (It isn’t) This article is not meant to show that GameGate is mostly about harassment. (It is) The sole purpose of this article is to highlight JW’s complete mathematical incompetence to such a degree that anyone will think thrice before using him as a source of numerical-based information again. Full-disclosure for the ethics fetishists; I don’t like JW very much and consider him to be amongst the most vinegary of douches. But unlike JW, I can actually math, and these insults are merely a way turn my catharsis into entertainment.
Numbers don’t lie, but JW does!
JW’s easiest to spot error tellingly occurs the first time he does a mathematical operation more complicated than counting. Here it is:
For those that didn’t notice or can’t read the smallish print, JW says 1.8 million tweets over 30 days averages out to 30,000 tweets per day. 4th grade arithmetic, however, says this average should be 60,000 tweets per day. Even worse is when this error is pointed out to JW in the comments; JW justifies using this objectively incorrect calculation by claiming it’s for “wiggle room.” Statistics, by their very nature, offer plenty of wiggle room even if you are honest about your calculations. If you need to purposefully make miscalculations in order to get a desired outcome, you are truly a shit statistician. Numbers don’t lie, but here we have proof that JW will and does.
The scatistician that is JW goes on to abuse the Arabic number system in the very next paragraph. Here, he says:
Did you catch it? 10% of 1.1 million is 110,000. The 124,076 tweets aimed at Sarkeesian et. al. are more than 110,000. How the fuck does this happen? Seriously, how does a person, countering an article whose ultimate concern is ‘which number is greater’ completely fail at making a statement about which number is greater!? If a person can’t be trusted to do elementary math, can you really trust their insights on inferential statistics? (No, you can’t)
Does JW possess the knowledge needed for such an analysis?
At this rate, this article will become way too long if I address every single JW “miscalculation”, so I’ll skip ahead to his appendix, which is where he kicks his train wreck of thought into full speed! Here, JW serves a goulash of bad assumptions, faulty comprehension, and poorly drawn conclusions, all served with a healthy topping of pretentiousness, which is nothing short of him vomiting magic. How can you describe a person who, amidst a statistical analysis concerning gendered harassment v. ethical journalism, digresses to ponder the HIV/AIDS epidemic in sub-Saharan Africa? That is a question I cannot, nay, I DARE not answer. A question I can answer is, “Does JW possess the collegiate-level knowledge of statistics needed for such an analysis?” (No, he doesn’t)
Pseudo-intellectual sprinkles on a bullshit sundae.
Thankfully, the focused idiocy below adequately summarizes the masturbatory ravings strewn throughout his appendix.
Simply put, this is some Grade-A bovine manure concentrate right here! Quickly:
1. “Normalcy” isn’t really a word a statistician would use when describing a normal distribution, “normality” is. I’m pretty sure JW includes the terms “normalcy”, “Kolmogorov-Smirnov”, and “Kruskal-Wallis” just to put pseudo-intellectual sprinkles on his bullshit sundae.
2. A lack of normality doesn’t rule out all parametric models. That’s like saying if something is not an apple, then it automatically is not a fruit.
3. These stats are grouped by target and sentiment, not dates. Knowledge of the exact dates is completely irrelevant.
4. Common sense would suggest these groups are not independent, as Newsweek specifically chose them due to their connection to #GamerGate. But rather than follow in JW’s missteps and accept an assumption as fact, we could test the assumption using a statistical tool such as…
…a chi-squared test! I swear I heard angels singing when I read this, because a shitty statistician announcing they are going to use a chi-squared test is like an old man starting a statement with, “Not to be racist, but…” You don’t know exactly what’s going to come next, but you do know it’s going to be a marvelous display of ignorance. I could not wait to see JW defile a chi-squared test, which he inevitably does, here:
If you’re not a statistician, this probably doesn’t seem like a big deal. If you are a statistician, you might be doing quite the facepalm by now. To put it in layman’s terms, JW using a chi-squared test to calculate an expected tweet distribution would be equivalent to a chef using a knife to bake a cake. JW is spouting absolute gibberish here as he claims to have used a tool in a way it literally cannot be used. What JW Bonehead McGee actually does is calculate the above expected values based on the ridiculous assumption that the myriad tweets grouped by target and sentiment are independent. He then fumbles his way into performing a chi-squared test correctly, only to immediately re-fuck himself by misinterpreting the result.
The “(p<0.001)” result doesn’t validate any conclusion other than, “your assumption that these groups are independent is a pretty fucking bad assumption, and these groups are actually dependent on something.” A real statistician would use more statistical analysis to identify what that “something” is. But JW decides to mislabel proof of his incorrect assumption as proof that “women are more likely to receive more positive or neutral tweets.” Are these six people enough to make a reasonable conclusion about of Gamergate’s overall attitude towards gender? Considering JW’s opening salvo doubted Newsweek’s selection of 25% of #GamerGate tweets for their study, he definitely seems to be aware that conclusions based on questionable sample selection are unreliable. So of course, JW confidently pole vaults from a sample size of 6 people, i.e 0.0000030% of Twitters 232 million active users, to a sweeping proclamation of GamerGate’s gender politics. Ironically, a chi-squared test would have been a perfect way to determine if these six people are a good sample on which to base such conclusions. (They aren’t)
Speaking of conclusions…
It should now be obvious to any honest person with a few functional synapses that JW’s “rebuttal” is really just techno-babble bullshit spread over a parametric curve. It should be obvious that GamerGate’s statistical emperor has no clothes. It should be obvious that anyone who used JW’s analysis as reliable source (I’m looking at you @CathyYoung63) should be ashamed. But none should be any more ashamed than JW, the great deceiver himself. JW is a mathematical charlatan. JW is to numbers what Shakespeare’s Dogberry is to the English language. JW’s baseless conclusions are less reliable than those of a lobotomized chicken. He is willing to use whatever mix of animus and ineptitude is necessary to perpetuate his desired narrative. And while my own rebuttal may not be timely, that point is moot because JW’s analysis was, is, and always will be, a spectacular display of pure, unadulterated, numerical-based tom-fuckery.