Why we need to talk about Ron Schnell’s “charts” in the Johnny Depp v Amber Heard Defamation Trial
On Day 19 of the Johnny Depp and Amber Heard Defamation Trial an “expert” witness in statistical analysis, Ron Schnell, gave his expert testimony.
I won’t repeat how extraordinary he is as a programmer. A Google search will reveal that he began lecturing in college at the age of 14 and has a series of humbling accomplishments — each of which is probably unachievable in the ordinary person’s lifetime.
But his statistics suck balls.
He makes textbook mistakes and does a worse-than terrible job at statistical analysis and reminds us why Correlation Is Not Causation.
Correlation Is Not Causation
We’re used to causes correlating with their effects.
You push the cup, it tips over the counter, it falls, it smashes, and its contents sink into the grooves of your kitchen tiles.
In this case, Schnell finds a correlation between mentions of Waldman and negative tweets about Amber Heard.
And here’s the fundamental flaw.
He’s ignoring everything else — including other news coverage, evidence made public, and media related to the case.
Is Schnell correct and it is purely by the merits of Waldman’s effortless charisma that people are making negative tweets about Amber Heard?
Or could they be motivated by something else — like seeing video evidence showing that she has been grossly dishonest in her testimonies?
I’ve been manipulating images for around 25 years (for fun) and have proven my proficiency in writing image processing algorithms at Queen Mary University of London.
I’ll eat a bowl of nail clippings if one of those images is not produced by applying filter(s) to the other every day for a month.
According to Schnell, seeing convincing evidence that Amber Heard is lying could not possibly be a cause for people having a negative opinion of Amber Heard.
Amber Heard persists to lie even when presented with irrefutable video or audio evidence in court.
People do not like that.
But according to Schnell, they absolutely could not be factors at all.
According to Schnell, people could not be making references to Waldman because he has eloquently communicated what they already believed to be true based on video and audio evidence showing Amber Heard was lying.
Initially, videos surfaced showing Depp acting violently towards cupboards.
I was apprehensive about siding with Heard because I wanted to see the video in context as I know how abusers are able to orchestrate scenes like this.
The rest of the video confirmed my suspicions. She antagonised him.
She was not a victim. She does not fear Johnny Depp.
She berates him for complaining about her “decking” him in his jaw.
But according to Ron Schnell — none of this could possibly have caused people to have a negative opinion about Amber Heard.
I’m hoping Jeremy Howard or Rachel Thomas from Fast.ai will further elaborate on this as I find it concerning that a man with such a poor grasp of statistical analysis was allowed to present himself as an expert in statistics on the merit of his genius programming.
I’m a programmer. I know how easy it is to write programs that can perform advanced methods of statistical analysis that a statistician might find difficult (or nigh impossible) to write.
But that doesn’t make one immune to making rookie errors in statistical analysis.
What’s worse is that his supposed expert authority may mislead a jury, while the field typically requires sufficient skill to recognize these fatal sins.
Expert witnesses should be subject to a panel of field experts who can give unbiased guidance or corrections in cases like this when the “expert” clearly has overlooked the most basic and fundamental matters.
An expert witness can do far more harm in misleading the jury than a series of masterful liars on the stand.
Not a Valid Root Cause Analysis
There are much more appropriate methods for identifying whether Waldman is the sole cause for people’s negative opinions of Amber Heard.
- Find a group of people not yet exposed to the details of the case.
- Present items relating to the case but in different orders and measures (i.e. some people will never be presented with Waldman’s comments).
- Measure changes in their opinion after being exposed to each item.
This is how Facebook, Google, YouTube, Twitter, Amazon and friends figure out which adverts, videos and other content to give to you.
This is also how you can much more accurately measure the impact of a potential cause — all of which seems to Schnell fails to consider or take into account.
