Read this paper by a researcher with 100s of citations

And now imagine the amount of mistakes and stupid theories in theoretical physics works. Nobody bother to check all that shit.

I actually know at least one example of a "perspective theory" with 100+ articles about it which was debunked multiple times but nobody fucking cares.

Do this.

I got several early career papers as well as many good contacts by doing something similar, but in a smart way.

I'd go through publications, find inaccuracies and overlooked issues, address them, make a paper-draft-masquerading-as-a-report about it, contact the original paper's PI about it, publish that together.

Smart move, but which field of research are you in exactly?

Unless you have unrestricted access to any kind of lab equipment you want from raman spectroscopy to particle accelerators, Its generally not possible to do what you are doing if many researchers are involved in the published paper you are trying to scrutinize.

Also would your supervisor approve of you trying to repeat other researchers experiments instead of doing what you are funded to do? I know my didn't.

Probably not in your field but I often find that often methods sections are under specified. Not only purpose I jut think people do a good enough job of writing it for people that aren't them.

This is quite normal now, now one really reads most of the papers, so people, when in need just squeeze out some meek, sometimes partly made up papers.
>Physics friend
>Doing research, checking out this quite obscure paper about optics/something/microsopy/polarisation/etc
>Using formulas in the paper on data
>Nonsensical outputs
>Hits up author of the paper
>Ayy lmao, that method doesn't really work
>wat

Seeing as you're Veeky Forums and hundreds of other real scientists have gotten it right, you're probably a brainlet making a Gen Chem tier mistake

You're probably doing something wrong but don't realize it.

The Dunning-Kruger is real.

Nice strawman, though no doubt a classic tactic to throw away investigators from looking into potential research fraud

It's probably not that uncommon. There was a nature journal survey found like 70% of scientists had failed to replicate. In a lab I was working in I seen too albeit in neuroscience where they failed to replicate their own experiment several times when tying to expand on the original and decided to eventually stop coz was relatively costly.

>appeal to accomplishment

I'm a chemist and from my experience, very much that is published is literal trash.
Especially indian and chinese papers, if not from a top-notch journal, are almost always unreproducible trash.

Example: 5 groups did a reaction with a rather low yield (like 30-50%), but then there's one indian paper published in the indian journal of chemistry (or, very often, tetrahedron letters), that says 'no solvent used' and has a yield of 96%. I swear to god I've encountered this so many times that it physically hurts.

Even well known researchers with thousands of citations tend to cheat on their papers and write down higher yields than they actually obtained.

Example I heard: Corey (nobel prize winner) told his phd student "you tried everything and got 30% yield, a post-doc would do 50% and I'd get 80%, so we write that in the paper". i don't remember the exact quote, but it went something like that.

Hell, even in some groups of my university i've encountered people saying "that reaction only worked when X did it" and then person X was very smug about it and had his/her little tricks you first had to learn. this shit happens when you don't write the little important details, which are only very rarely documented in papers.

>40% of statistics are made up

Got a source on that? Thats a serious accusation against Elias James Corey. Sure, some students commited suicide while he was supervisor, but I don't believe he's ever been suspected of advocating or engaged in research fraud.

It is not fraud, it is, just, you know, alternative data

Talk shit about him on pubpeer

and technically, the alternative data can be reproduced, you just need to be in the right mindset, man

a prof at my university used to say this all the time (not a reliable source though, but whatever)

a friend of mine did a reaction from a corey paper that said 80% yield.
he run the reaction like 20 times, looked for every detail, but got only 5%-35% yield. he wrote an email to the author and never got an answer.

also, google the Corey 1,2-rearrangement (when you yield is not good enough, 19% becomes 91%).

I'm pretty sure your friend is just a brainlet who is doing the procedure wrong somewhere.

woah, is this why my chemist gf is always so stressed out? could it be that her yields are in fact not the problem, but the procedure itself is corrupt?

No, women in general are just overemotional and incompetent

This thread sounds highly suspicious, because careers are ruined when people are caught lying.

Are you trying to replicate methods in really shit papers? If they are high profile papers and the methods still don't work, then why don't you cash in?

Demonstrating a high profile paper is inaccurate sounds great.

lol she gets the highest yields in the class every time though m8. although that probably says more about higher ed than it does abt her

YES. Never assume you're the one who's wrong because they are published. If you fall on your face whatever, but eventually you will bust something wide open

So how do we know if your statistic isn't made up?

I don't think it's as easy as you say. the scientific community seems to be an incestuous and protective bunch that will defend their theories to the bitter end no matter how much fudging said theories might incorporate

I'm 83,21% sure it's right.

>read about Bruce Bueno de Mesquita predicting political events
>dig up his papers, trying to recreate his model
>fail
>welp, i'm not a smart person
>read a paper from RAND corporation recreating said model
>they failed, had to plug in his own math to make it work

and the guy published like a dozen of papers about it

no code = no publication

rand.org/content/dam/rand/pubs/rgs_dissertations/2011/RAND_RGSD290.pdf

1/10 people give a shit

I think it's really hard to proof a method or other results wrong in some fields, because you'd have to have bulletproof evidence.

You'd also have to put in much effort and time just to gain said evidence.
In reality, the time you'd spend proving someone wrong would be the time you'd otherwise use for your research.

So, naturally, people just say "meh that paper is shit", because otherwise they'd spend ages to proof someone wrong. (Even if it's a high profile paper)

In the worst case, you accusation might just be dismissed by something like and people would say "That guy just wants some attention because his research is not very relevant".

...

In all honestly you probably did something wrong.

Chemistry is indeed the most unreproducible.

This is why maths > science

I'm was (and still am) in bioinformatics and have a computer science background.

Basically during my PhD I was often doing busy work re-implementing existing tools and pipelines in order to a) understand the algorithms properly and b) make a non-shitty (or just adjusted for our type of data) implementation of something the lab needed, since most academic software is a proof-of-concept mess.

A few times I essentially stumbled on these big errors, hence the very productive result.

One re-implementation ended up licensed later on too giving me a small amount of income.

But yeah it's easier in a field like mine where all it took was processing power which we've had an abundance of.

ncbi.nlm.nih.gov/pubmed/15604363

this paper was published in 2005, retracted in 2007, but continues to be cited in 2017.

> physics and engineering
why the fuck are those put together?

should have copied your PI on the email

>that says 'no solvent used' and has a yield of 96%
lol, i thought we had it bad in biology

>Be microbiology student
>Working on a Bacillus genetic mechanism
>Trying to make a defined media published by a US Army scientist
>No evidence of growth
>Double check I followed the list of chemicals in media in methods
>Wtf man, gonna go contact the researcher for more info like what B. anthracis strain he used, how long was the media left, etc.
>Go to contact author, his name seems familiar
>Turns out he was the real 2001 anthrax killer who killed himself before they could prosecute him
>Mfw Evidence of bacterial growth the next day

saved

replication crisis

Pre-register.

Have a third party examine hypotheses and plans for data analysis before performing experiments. This prevents scientists from discarding what they don't like, and picking the results that they do.

Results are much more likely to be reproduced, and if they cannot, it was junk. It is much easier to separate the chaff from the wheat, which is much more important in this age where anyone can publish.

Why not? You couldn't reproduce the results?

Because engineering research is fundamentally different than physics.
In my engineering field something like 95% of the results are reproducible because most of it's just data analysis/building shit.

1. It was a joke.
2. Engineering could have/should have been included in other.

>careers are ruined when people are caught lying.

This is true but you have your conclusions from it completely backwards. Scientists are overwhelmingly conflict avoidant, and not without reasons.

The problem is so ubiquitous and fostering good relations with other groups is so important that directly calling out anyone on anything is generally avoided at all costs in most fields I'm aware of. The degrees of walking on eggshells and ignoring elephants in the room in the scientist profession are staggering.

Scientists almost never openly catch anyone lying, if we absolutely have to call someone out at all it's always after consulting with him, often publishing about the issue together, always with an assumption that it was an honest mistake done in good faith. Even when that is obviously untrue.

Because refutations ALWAYS get published

If you want emotional answer:
>Yeah that stupid researcher was obviously wrong! Haha! You are the best bro! You uncovered a conspiracy theory! Yeah! Yeah!

If you want scientific answer:
>Ask your colleague to repeat the method
>If he/she can't make it work either, publish a paper refuting the method (together)
>If he/she can make it work, inspect your procedure very carefully

>95% reproducible
>implying that's not God tier
Welcome to Science. You'll find the "Introduction" department at the back of the hallway next to the cleaning closet.

>Chemistry is indeed the most unreproducible.
>graph literally says the opposite

>low tier science pleb complains about not getting reproducible data
>doesn't even consider that his facilities are different and that it makes YUGE differences
Some places of science are just built on ancient indian burial grounds or some shit, it just don't work.

>Captcha: saline

Can confirm

Recently changed labs and couldn't reproduce any of my results until I rubbed the benches with Hind's blood

Thanks for the scientific confirmation

Currently we're working on already established protocol (fetal liver reconstitution in mice) that not one of the hundreds of people at this facility have been able to accomplish.

>inb4 irradiation machine isn't callobrated

Check again. Chemistry is the worst.

That's not what says though

in which field is this?

If you still have problems, email the ones that cited him (if they used the method of the article) about how they got the results.

Also, did you send an email with a detail of what you did and where it started to fail?, that could help the author to understand wtf is going on

It has the narrowest bar on the graph. Or are you suggesting that nature has a muddled way of presenting the data that conclusions may not be reproducible?

Biomedical?

nature.com/news/reproducibility-crisis-blame-it-on-the-antibodies-1.17586

>40% of statistics are made up.

I never implied that. Someone during peer review will see your mistakes (if you're wrong) and correct it instead of resorting to ignoring you like the author in OP story.

>that pathologist who states that during educastion no-one ever told him to evaluate antibodies before use
Funny. Some years ago when I served on one of those labs, when I wanted to evaluate pcr protocols, antibodies, cultures for strict reproducibility, and my co workers mobbed and bullied me, ridiculed me, etc.
And knowing to evaluate reagents and conditions is not something you have to be taught imho, its something you should know on your own. Modern day mol bio is down the gutter, and mostscientists dont want to admit it. Instead they cry about Trump and that pampered Syrian refugee "scientists" cannot spend German government money on U.S. conferences.
The system is rotten.

Remember the 11 gorrilion goy!
What are you, a racist!?!?!
You don't want to be seen as a racist, now do you!!!!!?!?!?!!!?!!!!?!!

Papers are written and published to generate grant money, not to promote science.

...

I encourage this as well. Shitty researchers will publish anything without regard for the consiquences made on the community. This is the kind of shit that fuel antivaxers.

That joke is really old. Strange that Veeky Forumsentists haven't heard it before.

>most academic software is a proof-of-concept mess
This makes me mad at least twice a month, there are so many tools which run only on the provided examples and have so obvious bugs on everything else and than you want to use their tools and even the original programmer has no fucking idea how it works. And some tools just aren't installable or fail to compile unless you know the arcane gcc parameters you need to run this mess.

I remember that I wanted to compare several approaches and none of the 6 or 7 papers had a working implementation

probably newfags from /pol/ and/or reddit
80% of shiposters here come from there

>tfw I got warned for this
mods
mods pls
I was being facetious
pretty funny they took such an obvious shitpost so seriously

>do a project in fluid dynamics (some fancy smoke simulation and rendering)
>wonder how people programmed their solvers in papers I've read
>at last find the source of one (1) program
>mfw
And people here have claimed that mathematicians are good programmers. No, fuck you. Anything more than 1k lines and it becomes a huge unimaginable mess.

I've heard it many times before, but not in a few years which was enough to make me forget about it and when I was reminded of it I had a little smirk

Refutations never get published. You write a comment on the article instead. This is absolutely shit but that's how things go.