The Failure Of Twitter Scientists: A $SAVA Story

Introduction

Welcome back everyone! My name is Hector Garcia. I hold a masters in pure math, I teach college Stats at several campuses, and I am a trader. I can be found mainly on twitter at @BruinStocks

So previously I wrote a reply to Dr. Elizabeth Bik’s data work on PTau-181 biomarker data from Cassava Sciences’ 2021 AAIC poster presentation. I am here for a second round, this time…ALLEDGELY versus the government itself!

Let me explain.

The Claim

Dr Jesse Brodkin (twitter: @jesse_brodkin) claimed he had received data from a Freedom of Information Act request on Cassava Sciences, Inc. SAVA’s proprietary biomarker diagnostic SavaDX.

Dr. Brodkin’s post included a Sava Biomarker photo (from SAVAs original 2021 AAIC poster presentation) and overlaid what he claims to be missing data points that SAVA excluded,. The claim here is that they came upon this document via a FOIA request. Posted below for full transparency is the original picture presented by Dr. Brodkin:

Of course, being a $SAVA bull (despite not having a current position) I was skeptical of the information being shown. The very first thing that had me scratching my head: that ridiculous mean the PURPOSELY slided in on the 100mg Placebo group.

Using WebPlotDigitizer – the application Dr. Bik used to extract her PTau-181 data for her write-up ( https://apps.automeris.io/wpd/ ), I extracted the value of the mean change in biomarker levels as 86.71% average INCREASE (that’s BAAAD) in SAVAs biomarker that was tested.

Now this mean, looked goofy to me. I cannot imagine one outlier, even of the size of 350% dragging the mean that much! After all, the 100mg Simufilam group with the original 21 data points had 17 data points at around -50%, and the data the FOIA added included a near -100% reduction in this biomarker! If SAVA was cherry-picking data they missed including the most impressive data point. Weird…

I did my due diligence and busted out ol’ WebPlotAnalyzer and extracted all but ONE data point: the 350% outlier – this value was not posted to scale because it would clearly fall off the chart if this was to scale (the scale of the axis), but fear not…as you will see I most definitely added the 350 value in the data analysis. So I extracted all other data points for the 100mg group INCLUDING the ones added by the FOIA

This time I didn’t use any fancy R code to analyze. This was done Excel…Excel!!!!

As you can see, I did indeed put that 350% data point into the data set. You can run the same extraction method via Web Plot Analyzer, or really any data extraction technique you like, even eye ball it and see what you come up with! I welcome all questions about my method.

Some Thoughts and Speculations

So when I originally saw that mean, I immediately thought that it was just made up. Then one my good friends @ajacksonshorse pointed that the three data points looked like they would average out to something in the 80s. He did eye ball the three data points and hit it on the head when he calculated the mean as roughly 86. He did a quick eyeball calculation on the three and came up with a mean that was nearly exactly correct! So it turns out that the mean is real, but its not the mean of the ENTIRE 100mg group’s data.

This isn’t trivial. Not by a long shot. Real statistical work has to be deception free, it must be unambiguous. We were given NOTHING that implied the mean bars were supposed to be the mean of the 3 data points.

Naturally, as it should be: you cannot put a mean bar on a data set for which the mean doesn’t represent. NO ONE would do this IN GOOD FAITH. This is a classic example of an intentionally deceptive graphical display of data. This is awful, regardless of what the bears’ overall goal is (to show $SAVA is a fraud).

Its ok to use original documents, its even ok to question every piece of data. What is NOT OK, under any circumstance is purposely publishing a data plot with statistics that don’t reflect the data set it is IMPLIED to describe.

In the end, that’s the bottom line. If the person who created this graphic couldn’t be bothered to spell out exactly what that mean is supposed to represent, then they have no business posting material thats damaging to SAVA and damaging to SAVA investors, an “oversight like this” is no way defendable.

I am sorry Dr. Brodkin, you don’t get to wave this off as it being “so obvious.” Newsflash, this isn’t obvious to the majority of people. This is par for the course of the typical argument by @QCMFunds and their gang of scientist

Gang of Deceitful Scientist: There’s fraud! Believe us, we are scientists!

Our team: Hey this and that from your argument doesn’t make sense, could you explain?

Gang of Deceitful Scientist: “Just look at it, its obvious.”

Our Team: “Really? You expect this stuff to be obvious to people that don’t have a grad level chem, bio, pharm education?”

Gang of Deceitful Scientist: “You don’t need a science degree to see this is all obviously fraud.”

Our Team: “Really? Care to educate me on the science?”

Gang of Deceitful Scientist: “I don’t have time to explain things to you (but I do have the time to argue with you for months on end)”

Our Team: But this doesnt sound right!?

Gang of Deceitful Scientist: Well what are your credentials to argue against us?

Our Team: You already know we are biochem people.

Gang of Deceitful Scientist: “Well then! You should listen to real scientist, you dont know anything but we are trained!”

Such douchebaggery. But again, we are supposed to trust this cabal, and not the research scientist who have been doing this research directly, for a decade plus!

Sorry, I don’t appeal to authority, neither you nor SAVA, nor an Ivy League researcher. When I spot BS in data (and data is surely in my space of knowledge, maybe even more so that anyone in the cabal), I will call it out!

I was asked recently why, if I don’t hold a position, am I such a $SAVA fan. First of all I trade this stock so it isn’t always in my portfolio. Secondly these coordinated attacks on $SAVA have made for big surprises when I log into my platform an hour before the market opens.

But since I am not trading $SAVA until all of this is resolved, my main beef is not with the objections to SAVAs data: its important to question everything. No my beef is with the constant use of purposely deceptive tweets and especially purposely deceitful “scientific” papers are beyond ethically wrong.

So yeah, $SAVA might still not get approved, hell $SAVA may well be found guilty of fraud. In no way does this excuse the use of deceptive dissemination of scientific data.

I really don’t care if Dr Brodkin or Dr Adrian Heilbut or Dr Bik or @QCMFunds have an excuse or “explanation” for this, but there isnt one to be had. Seriously, there is NO RATIONAL AND ETHICAL reason put a mean on a plot where the mean isnt relevant to the entire set. Period. As distinguished researchers and scientist its not a matter of whether they should have known better, no, this is deception is so blatant that they indeed DO KNOW what they are posting, which makes this whole QCM led bear attack despicable

My Thoughts on Dr Brodkin’s Post

I am, in no way claiming that this is Dr. Brodkin’s work. In fact I know its not his work. I had made mention to Dr. Broadkin of my intentions to extract and analyze the data several times.

If he knew the mean line was averaging the 3 added data points, he would or should have told me with in this thread. Even after this exchange I kept mentioning what I had found out about the mean, and still never got any information about what the mean bar was “supposed” to mean.

I showed him my eye-balled data set and he still had nothing to offer me in terms of what the mean meant:

It wasn’t till I announced on Twitter of my intention to publish what I had found that Dr. Brodkin responded. Amazingly he tries to sneek in the out:

So I do not think Dr. Brodkin created this picture.

Seriously, one more time: there is NO REASON, NONE, that would compel a scientist to slap on a statistic on a data set that isn’t represented by that statistic. To do so, and not to flat out spell out why you are putting a statistic on a part of the graph where it does not belong is FLAT OUT DECEPTIVE.

Whoever created this document INTENDED TO DECIEVE anyone who appeals to authority and just buy this because “an expert said so.” I will not be shaken from that position either. This type of basic ass deception absolutely sickens me, no joke, I teach stats mainly for the sake of giving students some tools the can actually use, regardless of their major. This behavior is the absolute antithesis of what I stand for. This is not a minor mistake, this isnt a minor blunder, and this is not a misunderstanding. There is zero excuse for those mean bars to be placed where they are, NONE.

This is clear, deliberate deception. And no Dr Brodkin, you cannot dismiss this the same way you’ve tried to dismiss my arguments on Twitter. I understand this is an embarassing situation, but I am sorry, no you can’t pop up 3 hours later and suddenly decide to act like “its so obvious that mean was the mean of only 3 data points. No, I am sorry, but that doesn’t fly in ay airspace.

I do challenge you to present any other data plot, from any reputable journal, on any subject and show me an example of a plot that included a statistic into the plot of a data set which the statistic didn’t represent WITHOUT EXPLICITLY stating what the stat was referring to.

Hector Garcia

@BruinStocks – Twitter

Latest Developments

This section is meant to continue to update readers on the latest developments. I don’t want to walk back anything I’ve posted and want everyone to know when something is new to the article.

Epilogue 1 – 12-01-21 – 7:30APST – So we have finally figured out exactly what the “extra points” are and why they weren’t included, and to be honest I should have remembered this detail: ClinicalTrials.gov did state that 3 people dropped out of the experiment: two for not having any simufilam in their system, and one for not taking the prescribed amount of medication.

The original SavaDX Biomarker plot has 3 extra points. The number of people who began in the group was 21. After 3 dropped out, there should only be 18 data points, but the plot clearly shows 21 points of data. However SAVA has addressed this issue long ago: the plot has mistakes, but those mistakes did not carry over to the data analysis. There is record of Remi stating this:

So based on the fact we know the missing data points belongs to drop outs and SAVA acknowledging 3 of the data points are incorrectly inserted into the plot, what we SHOULD look to do is eliminate the 3 “extra data points” and 3 points from the original SAVA plot. Which three points? Well as our shorts have indicated, there is no patient matching info that would tell us which of the 3 points should be eliminated. So just for the sake of illustration, I’ll take out the 3 best original data points (i.e. the 3 most “negative” percentages), and this is how it pans out:

Upon doing this, I saw 3 data points that were repeated. This confirms that the 3 extra points the bears reference are not different data points for the same patient. I identified duplicate data on SAVA’s original poster (again, Remi has addressed this). If the bears really were about full transparency they would have done the basic data extraction and the basic stats I used to confirm their accusations. They did not. For what its worth, the group has smeared Dr. Bordey’s work:

https://pubmed.ncbi.nlm.nih.gov/32075941/

And have harassed her about not performing FilA binding experiments herself. If this group expects Dr. Bordey to do her DD, then they should also do the basic DD to fully back up their claims.

Again, this is an INTENTIONAL ATTEMPT TO DECEIVE and I am flat out disgusted by this.

Hector Garcia

@Bruinstocks – Twitter.
Disclaimer: The author has no position in SAVA, but may enter a long position in the next 72 hours.