[PmbPhy (OP)]

Many years ago a little book came out called How to Lie with Statistics by Darrell Huff. It's available online at: faculty.neu.edu.cn/cc/zhangyf/papers/How-to-Lie-with-Statistics.pdf

If you don't have the time and/or inclination to read it then here's a webpage on the subject
https://www.psychologytoday.com/us/blog/the-fair-society/201111/how-lie-statistics

It was such a good book and so well received that it's required reading it some courses on statistics. By that title the author is not suggesting that people lie with statistics but that in some cases use them improperly etc. The details being described in the book of course. Its an enjoyable book to read too.

I started this thread to ask when we can trust the statistical information provided to us. And it comes in many forms. Are any of you familiar with D. Oz? He's on TV and in magazines a great deal. He always seems to me to be pushing a diet something or 'miracle' diet, green coffee beans and the like. When people do that they're really saying that something has undergone scientific testing which showed that something had a given property. He had to appear before a congressional hearing on all of this. For details please see: CNN report:

Congressional hearing investigates Dr. Oz 'miracle' weight loss claims
https://www.cnn.com/2014/06/17/health/senate-grills-dr-oz/index.html

Question: Does Dr. Oz lie with statistics?

As many of you know, I'm deeply concerned with putting the US opioid problem into proper perspective. I was going to dig into the details of the stats used in the medial data on this subject so as not to be unscientific but that's a daunting task. Consider this example

The Effectiveness and Risks of Long-Term Opioid Treatment of Chronic Pain at
https://www.ncbi.nlm.nih.gov/books/NBK258809/

They reference 4,209 citations. There's too much data to look at and that's for one paper to consider.

Its for reasons like this I decided yesterday to alter my approach. First, rather than concern myself with whether its a legitimate crisis or epidemic or not I'm merely going to address the concerns associated with it such as the amount  of danger its putting patients in and the amount of abuse they are subjected to. Those are documented to some extent and for what isn't documented I'll use myself as case study. That means exposing my health problems to the public. Blech!

As a concrete example of misuse of statistics which may wrongly influence the public is online at the CVS website.  One might imagine that people who can influence the public know enough to be cautious about the use of such things. When it came to one of the presidents at CVS pharmacy that turned out to false, and in the worst possible way. Last year I came across an article in Morning Consult which appeared on July 11, 2017 entitled

Time to Put Limits on Opioid Prescribing by Larry Merlot, Larry J. Merlo, President and Chief Executive Officer, CVS Health. Thankfully for us CVS placed it online at their website at
https://cvshealth.com/thought-leadership/time-to-put-limits-on-opioid-prescribing

Merlot used a public opinion survey which is also online at
https://cvshealth.com/sites/default/files/cvs-health-prescription-drug-abuse-and-disposal-report.pdf

I'll leave it at that for now. I want readers of this thread to experience that misuse for themselves.

[PmbPhy (OP)]

Regarding my comment on the amount of danger and abuse I mentioned. Last year I got hit by a car as I was walking through a parking lot. The poor gal didn't see me because I was in her blind spot. She felt so guilty she started crying. There was no way I'd sue so I wanted to walk away and forget about it. Onlookers who saw it kept asking me if I was okay. Slowly the pain started coming. It wasn't immediate but it was unsettling. So I agreed to go to the hospital to be safe. When I got there the pain was severe. Often a shot of toradol relieves it. This time it didn't. They refused to do anything else claiming a new law prohibits narcotics in the ER for chronic pain. I was shot with the arrow of binary thinking. It wasn't chronic pain alone. I was hit by a freaking car damn it!

But you can see the total absence of empathy in their expression. Its replaced by contempt.

The other day was different though. I went to my local ER for my server pain from sciatica and they gave me a shot of that and it worked fine.

[PmbPhy (OP)]

Is anybody here familiar with the quote

"The only thing necessary for the triumph of evil is for good men to do nothing." - Edmund Burke

If so then have you ever applied it in your life?

[PmbPhy (OP)]

One of the things I planned on telling my gov reps is how to save a lot of money. Right now a lot of people in crisis call an ambulance to take them to the ER. That costs thousands of dollars. Many times its because they/we can't afford a cab.
Taking the bus would mean a delay of a few hours due to how the schedules run. And a bus ride that long is very painful and the bus ride back would produce the pain we went their to escape. To save all that money the state could act to create an ER transport service. It could be used when timing is important but the problem is not dangerous. That's much cheaper to do. I doubt it'd be more than a taxi fare. Or the medical insurance could be used to pay for fare to the ER. But it can't be through reimbursement of course.

I sent that to my state rep a few months ago and heard nothing back. That's a problem in itself.

[Bored chemist]

To save all that money the state could act to create an ER transport service.

It's not exactly a radical idea.
https://www.nhs.uk/nhsengland/aboutnhsservices/emergencyandurgentcareservices/pages/ambulanceservices.aspx

Except in the US, where suggesting it will have you hanged- twice- for being a communist.

[chiralSPO]

Regarding "When can we trust statistics?"

I would say that stats should only be trusted when they come with full disclosures of how the data were collected and analyzed, including confidence levels and explicit identification of the limitations of the conclusions drawn. It is best if there are multiple independent studies that support the same conclusions (especially if they have different/complementary limitations).

Unfortunately peer review alone is not sufficient to guarantee veracity, but it is a good start.

Most non-scientific media will not provide much of the disclosures mentioned above, but hopefully they do link to the original studies. Be extremely wary of overly simplistic claims, or studies that indicate high statistical significance of a very subtle difference (it is fairly easy to increase the calculated certainty without actually increasing the power (https://en.wikipedia.org/wiki/Statistical_power ) of the study).

The hardest flawed studies/reports to spot are the ones with a confounding factor that is not well controlled for or some type of bias in the design of the study (either intentional or not).

[alancalverd]

Statistics generally are more trustworthy than the inferences drawn from them.

Recent shock horror headlines announced a huge disparity between male and female pay at EasyJet. Company representative said the statistics were absolutely accurate and entirely acceptable: "We pay pilots more than cabin crew - that's the market. It just happens that most pilots are male and most cabin crew are female, because that's what people choose to train for."

[Bored chemist]

When Can We Trust Statistics?
When we understand them.