Monday, January 16, 2012

All data is wrong

Some data are useful.

Discuss

UPDATE: Fixed the grammar

18 comments:

Former Skeptic said...

*Winces*

Back when I was a student, I kept writing (and saying) that "the data is xyz...the data shows that etc...".

After one class presentation littered with such utterances, a visiting professor (and a NAS fellow) pulled me over to one side and told me in no uncertain terms:

"The data ARE, young'un. Only uneducated folks say otherwise."

Dallas said...

After twenty years of HVAC test and balance I can confirm that all data is incorrect as in inaccurate. Temperature readings with thermometers certified to accuracy of +/-0.1 degrees in a field situation may have errors of +/- 5 degrees due to a variety of reasons including stratification, condensate misting, separation, parallax and operator brain farts. The same with velocity pressure, static pressure and gas concentration measurements. At least two, preferably more methods to verify readings were needed to have a reasonable confidence of +/-10% of air volume and capacity calculations.

My rule was, "You know the data will lie to you, try to figure out how bad it is lying."

That makes all data useful, since they tend to rat on each other. :)

Martin Vermeer said...

> That makes all data useful, since they tend to rat on each other

So true, so true... must remember this

Russell said...

When no datum is correct all the data are wrong, so no data is correct, as there are no data when every datum is wrong; right ?

Martin Vermeer said...

Yep. And even a ticking, adjusted clock will only show the right time on a subset of the time line of measure zero

Hank Roberts said...

but ... but ... you can _graph_ the numbers!!

http://1.bp.blogspot.com/-iycWpiuCmvM/To5_e9CBzQI/AAAAAAAABqk/kwujRGwsAXY/s1600/FIGURE6-4-650.JPG
http://geo-engineering.blogspot.com/2011/10/thermal-expansion-of-earths-crust.html
"Thermal Expansion of the Earth's Crust Necessitates Geo-engineering"

With a tip of the tinfoil-lined hat to "climateforce.net" for the link

KAP said...

I assume this is a response to the old saying, "no models are correct; some are useful."

I have a better response to that saying: Every scientific law is a model. If you don't like models, you don't like science.

Jeffrey Davis said...

Worse. Each item can be the case or not the case while everything else remains the same.

Kevin O'Neill said...

I like to think that every data set tells a story. I keep some around for nostalgic reasons - especially the particularly humbling ones. I.e., Egads, how much time did I spend trying to tease something useful out of *that* before I realized the whole idea was nonsense?"

Marion Delgado said...
This comment has been removed by the author.
Marion Delgado said...

Harrumph. This, to me, is like Tobis' agreeing that peer review is broken. There's a point there, but it's one denialists tend to grab and run with.

I suppose one example would be Erwin Chargoff's DNA chemical analyses where the paired amino acids were in a ratio "close to 1" but he wouldn't say it was 1. But it was later shown that it was 1, so in that sense the measurements, the data, were wrong.

But I don't always agree you always have a better yardstick than the data by which to say all data is wrong. So I think that trope is wronger than all data. :)

Anonymous said...

Well Patrick Michaels has no issue deleting data that doesn't fit his PR campaign.

EliRabett said...

Dear Marion, surely you know that Eli is not nearly as nice as Michael, point being the next time somebunny tells you that they only believe the data and not the models . . .

Ian Forrester said...

Marion it is the nitrogenous bases which pair up, adenine with thymidine and guanine with cytosine.

Also it is Erwin Chargaff.

KAP said...

UPDATE: Fixed the grammar?

CORRECTION: you only fixed half the grammar! The title is still wrong.

Marion Delgado said...

thanks Ian - I was trying to somewhat agree - the point is that Chargaff's data was helpful but in the sense that he didn't say it was a 1-1 pairing, wrong. But I still don't think this is always true because it implies you always have a yardstick by which the data is wrong, and I don't think you do.

EliRabett said...

1. Evidently you have to be logged in to Gmail or the equivalent to post

2. The Grammar Police are invited to follow the link.

Holly Stick said...

I had to sign out of google and then sign back in for it to work. This time it worked alright.