[ctrl-j]

> When the models didn't show this it was presumed that the models were wrong so hyper-parameters were tuned to try to torture the models to produce the 'right' result. When even that didn't work they went so far as future pollute the data.

Please name and shame. I've also done work in the field, and this kind of behavior was never acceptable - and should not be.

If you let the bad actors hide in obscurity, they will continue to practice bad science.

[huffpopo]

Like a #metoo movement for scientists. AFAIK Hyper-parameter tuning is fairly normal and necessary given how easy it is for the models to become unstable on their own and future polluting is often accidental. It can come down to judgement calls and difference of opinions as to what degree of intervention is acceptable which leaves a lot of wiggle room If you removed all future pollution and and left hyperparameters at the initial best guess you often end up with crazy results in backtesting. E.g. if our past predictions were correct then we should be living on Venus right now which obviously we're not so we must need different hyperparameters. My former colleagues were trying their best and even they knew their adjustments were going beyond reasonable. They never signed off on the final results so the project was politically useless and subsequently killed. They've all since left the climate change industry disillusioned as well.

[Oletros]

> . if our past predictions were correct then we should be living on Venus right now which obviously we're not so we must need different hyperparameters

I would like to know hat models showed runaway warming and were publishing because I don't remember anything like that.

[huffpopo]

I was talking about backtesting of the models to see if they would have predicted the past correctly. I.e. You run the model from 1970 to today and measure how close the model prediction is with the actual measurements. If it's too far off then obviously the model is wrong and changes are made to try to improve it.

[Oletros]

> It was very important to the other researches that the models produce the 'right' predictions which at the time was runaway global warming. When the models didn't show this it was presumed that the models were wrong

Sorry, you have claimed that the scientists wanted the 'right' predictions, that was runaway warming.

I repeat, can you link to any scientists saying in 2006-2007 that the most probable outcome was runaway warming?

[balthasar]

What is the climate change industry?

[praxulus]

Wait, so the terrible science you're talking about was never actually published, yet you still distrust the rest of the field?

[huffpopo]

The group leaders were pushed things too far for me to accept as legitimate but not far enough to produce the results needed. We were aware of others that pushed things even further and they got to continue working. This adds a selection criteria bias to those still doing the research which has me trusting them even less. And this was before ClimateGate which made the shenanigans done by other groups public.

[gech]

Your reference to and unreserved use of the term "ClimateGate" makes me question your agenda here. Wasn't "ClimateGate" determined to be a smear campaign? I'm asking rhetorically as I can see the discussion devolving.

[todd8]

I generally keep my eyes open for climate change news and didn't see this. Do you have any reasonable references that explain that ClimateGate was a smear campaign? I thought it was a small number of committed researchers making some mistakes and straying from strict scientific protocols; I never saw that it was a smear campaign.

[Oletros]

https://en.wikipedia.org/wiki/Climatic_Research_Unit_email_c...