"The temperature record is unreliable!"

jwoodward48 wrote:I don't have any models on hand. I'm not the source for that.

jwoodward48 wrote: I agree that you do not have the burden of proof for showing that the models are falsifiable.

jwoodward48 wrote: But you do have the burden of proof for showing that climate scientists have fabricated data and used random number generators as models.

IBdaMann wrote:

jwoodward48 wrote:I don't have any models on hand. I'm not the source for that.

That's right. You take all that on blind faith. I had momentarily forgotten.

jwoodward48 wrote: I agree that you do not have the burden of proof for showing that the models are falsifiable.

No one does. Any model itself either is or is not falsifiable. It's patently obvious. Only the model speaks for the model.

One of my caution flags gets raised anytime someone starts using the words "The model says ..."

Another caution flag gets raised whenever someone uses the words "Model X can be falsified by..."

Nobody gets to say what a model says. Only the model gets to say what it says.

jwoodward48 wrote: But you do have the burden of proof for showing that climate scientists have fabricated data and used random number generators as models.

This is going to be an excellent learning opportunity for you.

Watch me not support my statement in any way. Yes, I claim that warmizombie activists fill the internet with fabricated data.

Question: Why might it be the case that I am under no obligation to support that claim?

Answer: I'm not pushing any religion. I don't care whether you believe that warmizombie activists fill the internet with fabricated data.

You, on the other hand, are pushing your religion and are the one making affirmative claims. The burden of proof for those claims rests entirely with you.

You well know how religions are. They need for people to accept their dogma and to join the church. You care about others accepting your dogma. Hence you are the salesman and you need to sell/convince your target market of current non-believers on your dogma. They don't somehow need to prove anything to you.

I, on the other hand, am happy if you're happy at abandoning your sales pitch.

If you're selling science then I'm buying. If you're selling religion then I think I'll pass.


* . r . * . e . * . l . * . i . * . g . * . i . * . o . * . n . *


.

Surface Detail wrote:

Into the Night wrote:

Surface Detail wrote:

Into the Night wrote:

Surface Detail wrote:

Into the Night wrote:

Surface Detail wrote:

Into the Night wrote:

Surface Detail wrote:

jerrylh wrote:
Hi IBda Mann
For my benefit, could you clarify why you feel that the temperature record is unreliable. Presently, the temperature is based on averages of several readings around the globe. Then an average is taken. Now that number by itself, doesn't mean much, but it is the changes that occur during decades that are meaningful. Are you saying that the readings are wrong?? If so, what do you attribute that to?? Are they readings being intentionally falsified?? I would just like you to clarify your position so I could better understand it. Thanks

I don't think you stand any more chance of getting any sense out of IBdaMann than the rest of us do. Go on, though, ask him for any evidence to support his contentions and watch the spittle fly. 3, 2, 1...

Edit: P.S. Don't try to explain to him what a "confidence level" actually is. It's one of his trigger words.

Don't try to use big words. You don't know the difference between confidence level, confidence interval, and margin of error.

Actually, I do. You might want want to have a quiet word with IBdaMann, though. He seems to think that "confidence level" is some sort of arty-farty IPCC invention rather than a statistical term. I'm sure you can put him straight.

Don't need to. He knows what a confidence level is. You don't.

The confidence level is the probability that a value lies within the margin of error. That's why a margin of error is meaningless unless you specify the confidence level. Obviously.

Wrong.

The confidence level is the probability that a value lies within the standard deviation, as typically built by a paired randR (that 'bell' curve you keep seeing).

That's not right. The probability that a value lies within one standard deviation is, by definition, 68%. The confidence level for a particular margin of error may, or may not, be one standard deviation. For example, margins of error are often given to confidence levels of 2 standard deviations, i.e. 95%.

Margin of error is calculated from confidence interval, which is calculated from confidence level. You are making a circular argument.

I'm not making an argument, circular or otherwise. I'm simply giving the usual definitions for these terms.

Surface Detail wrote:
The confidence level, expressed as a percentage or as a multiple of the standard deviation, is the probability that a measured value lies within the given confidence interval. If the confidence level is specified as 2 standard deviations, for example, then 95% of the measurements will lie within the confidence interval. The margin of error is usually defined as half of the width of the confidence interval.

Into the Night wrote:

Surface Detail wrote:

Into the Night wrote:

Surface Detail wrote:

Into the Night wrote:

Surface Detail wrote:

Into the Night wrote:

Surface Detail wrote:

Into the Night wrote:

Surface Detail wrote:

jerrylh wrote:
Hi IBda Mann
For my benefit, could you clarify why you feel that the temperature record is unreliable. Presently, the temperature is based on averages of several readings around the globe. Then an average is taken. Now that number by itself, doesn't mean much, but it is the changes that occur during decades that are meaningful. Are you saying that the readings are wrong?? If so, what do you attribute that to?? Are they readings being intentionally falsified?? I would just like you to clarify your position so I could better understand it. Thanks

I don't think you stand any more chance of getting any sense out of IBdaMann than the rest of us do. Go on, though, ask him for any evidence to support his contentions and watch the spittle fly. 3, 2, 1...

Edit: P.S. Don't try to explain to him what a "confidence level" actually is. It's one of his trigger words.

Don't try to use big words. You don't know the difference between confidence level, confidence interval, and margin of error.

Actually, I do. You might want want to have a quiet word with IBdaMann, though. He seems to think that "confidence level" is some sort of arty-farty IPCC invention rather than a statistical term. I'm sure you can put him straight.

Don't need to. He knows what a confidence level is. You don't.

The confidence level is the probability that a value lies within the margin of error. That's why a margin of error is meaningless unless you specify the confidence level. Obviously.

Wrong.

The confidence level is the probability that a value lies within the standard deviation, as typically built by a paired randR (that 'bell' curve you keep seeing).

That's not right. The probability that a value lies within one standard deviation is, by definition, 68%. The confidence level for a particular margin of error may, or may not, be one standard deviation. For example, margins of error are often given to confidence levels of 2 standard deviations, i.e. 95%.

Margin of error is calculated from confidence interval, which is calculated from confidence level. You are making a circular argument.

I'm not making an argument, circular or otherwise. I'm simply giving the usual definitions for these terms.

By making an argument, which happens to be circular.

Surface Detail wrote:
The confidence level, expressed as a percentage or as a multiple of the standard deviation, is the probability that a measured value lies within the given confidence interval. If the confidence level is specified as 2 standard deviations, for example, then 95% of the measurements will lie within the confidence interval. The margin of error is usually defined as half of the width of the confidence interval.

You don't get to arbitrarily assign margin of error or confidence interval. They are calculated against the paired randR equivalent.

jwoodward48 wrote:
That's not a circular argument.

jwoodward48 wrote:
That's not even an argument.

jwoodward48 wrote:
He's giving the standard definitions of those terms. We do get to arbitrarily assign the margin of error - but then there's a confidence level that goes with that. So I could say "0.001 margin of error... with 5% confidence". But that is usually utterly useless, so in practice everybody uses 90-99% confidence levels.

jwoodward48 wrote:
Furthermore, I'm not finding a single hit for randR online within statistics. For RNG, it's just "a random number between 0 and 1." Could you elucidate a bit more, or give some other related search terms so I can learn more?

Into the Night wrote:

jwoodward48 wrote:
That's not a circular argument.

Yes, it is.

jwoodward48 wrote:
That's not even an argument.

Yes, it is.

jwoodward48 wrote:
He's giving the standard definitions of those terms. We do get to arbitrarily assign the margin of error - but then there's a confidence level that goes with that. So I could say "0.001 margin of error... with 5% confidence". But that is usually utterly useless, so in practice everybody uses 90-99% confidence levels.

You get to pick against X (the paired randR), not the incoming data.

jwoodward48 wrote:
Furthermore, I'm not finding a single hit for randR online within statistics. For RNG, it's just "a random number between 0 and 1." Could you elucidate a bit more, or give some other related search terms so I can learn more?

The theory of random numbers is an entire branch of mathematics. I am not going to write a book for you.

Since you seem to have trouble finding information on it (Google is not God), I will elucidate a bit further.

A random number is any positive integer or zero. I am not going to bother explaining why at this point. The reason is quite involved, and it involves that 'domain' thing again, although in a different way. It also involves why mathematics is a closed system.

There are three basic forms for generating them, randU, randN, and randR. I am not going to bother explaining why they are called this at this point. The history of these names is long and buried deep in the world of information science.

randR is the basic random number. The same kind of random sequence you get from a die or from a roulette wheel. The number may repeat at any time, any number of times. A coin flip (with an official 'fair' coin) is randR.

Pairing two randR's creates the familiar bell curve. This is the same as using two dice and summing the total instead of using just one die.

randN is the random number sequence incorporating a memory. Once a number appears in the sequence, it will not appear again. This is the same kind of number you get from dealing a deck of cards, as in blackjack, poker, or War. randN inherently contains a moment of reset in the sequence (shuffling the cards, for example). Hence, randN is derived from randR through the use of an event.

randU is the fixed random number. The sequence is completely predictable. It too contains an inherent reset event, where the initial seed of predictability begins. This is the random number generator of a typical computer. It is useful for generating psuedo-random numbers and for extremely compact random data compression techniques.

The upside of randU is that you can put it in a computer program. The downside of randU is that the risk of predictability or even repeatable sequences occur. This can be alleviated somewhat by using a randR for the initial seed (the way a typical Linux machine does, for example). It also happens to be the method used in modern slot machines (ones newer than electromechanical reels).

You may consider each of these types of random number sequences a function, if that helps you to envision them. It isn't quite true, but it's good enough for now.

Surface Detail wrote:
IBdaMann is lying about the availability of raw data and misguided about the importance of sensor accuracy.

The raw and adjusted data are available to the public here, for example:

ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/

The margins of error for the thermometers are negligible compared to the errors resulting from positioning, surroundings and variations in reading time. It is to compensate for these errors that the data are subsequently adjusted using statistical methods.

jwoodward48 wrote:
The margin of error does not include either instrumental error or bias. It only accounts for the potential error induced from the sample being non-identical to the population.

jwoodward48 wrote:
But is an averaging method good enough to describe the average temperature of the region? Yes.

jwoodward48 wrote:
Why not? Keep in mind that the 95% MoE =/= the maximal error, and that the MoE for the average temperature for several regions is the average of the MoE for each of those regions (weighted by the weight of each region, naturally).

Into the Night wrote: You cannot use weighting. That violates the requirement of random selection. No data can substitute for missing data. Remember, the selection MUST be randN. Weighting the data effectively turns it into a randR.

IBdaMann wrote:

Into the Night wrote: You cannot use weighting. That violates the requirement of random selection. No data can substitute for missing data. Remember, the selection MUST be randN. Weighting the data effectively turns it into a randR.

Correct. "Weighting" is a polite word that warmizombies use to acknowledge that they have doctored/tweaked/cooked/fudged/skewed the raw data.

All unmodified raw data must be presented for scrutiny. Period.

All margins of error for all sensors/equipment involved must be presented. Period.

What is allowed, and encouraged, is to then also provide an analysis with weightings and explanations (also presented for scrutiny) if they make sense.

If I don't have all the unmodified raw data such that I can drop it all into a spreadsheet and run my own analyses then I have to ask what they are trying to hide. At what conclusions do they not want me to arrive such that they are providing me only modified data or worse, by only telling me the conclusions I am to believe without question?

Surface Detail wrote: I just pointed out where you could download the raw data from.

There, that's better. Was this another blank post by IBdaMann? Why does he do that?

IBdaMann wrote:

Into the Night wrote: You cannot use weighting. That violates the requirement of random selection. No data can substitute for missing data. Remember, the selection MUST be randN. Weighting the data effectively turns it into a randR.

Correct. "Weighting" is a polite word that warmizombies use to acknowledge that they have doctored/tweaked/cooked/fudged/skewed the raw data.

All unmodified raw data must be presented for scrutiny. Period.

All margins of error for all sensors/equipment involved must be presented. Period.

What is allowed, and encouraged, is to then also provide an analysis with weightings and explanations (also presented for scrutiny) if they make sense.

If I don't have all the unmodified raw data such that I can drop it all into a spreadsheet and run my own analyses then I have to ask what they are trying to hide. At what conclusions do they not want me to arrive such that they are providing me only modified data or worse, by only telling me the conclusions I am to believe without question?

.

IBdaMann wrote:

Surface Detail wrote: I just pointed out where you could download the raw data from.

Nope. You are simply weaseling again in your constant string of ongoing dishonesty. The story never changes.

1. I see your link and your promise of the Holy Grail of Global Warming data that shows everything you've been claiming. My search is over!

2. I go to your link and disappointingly I don't find the Holy Grail you promised. All I find is the following:

Mar 25 2008 00:00 10445 alaska-temperature-anomalies.txt
Mar 12 2008 00:00 3410 alaska-temperature-means.txt
Nov 10 2011 00:00 Directory anom
Sep 23 2016 13:03 Directory blended
Oct 17 2016 08:57 Directory daily
Dec 15 2009 00:00 Directory forts
Sep 11 2003 00:00 14953824 grid_gpcp_1979-2002.dat
Nov 30 2011 00:00 3886592 Lawrimore-ISTI-30Nov11.ppt
Feb 27 2013 00:00 Directory snow
Aug 22 2001 00:00 Directory v1
Oct 15 2016 09:00 Directory v2
Oct 17 2016 08:07 Directory v3
May 27 2016 09:03 Directory v4

3. I stand back and wait for you to pout like a fukcing baby and to scream "The data is THERE, you just have to search and hunt and scour and find it yourself! I'm not going to do your homework for you!"

4. I then tell you that you should do the legwork to go to all the places where you know the data to be, gather it, compile it yourself and put it into a consolidated Holy Grail of Global Warming datafile that I or anyone else can easily drop into a spreadsheet to perform analyses. If it's not important enough for you to do then it's not important enough for me either.

This brings us to the question of "Why hasn't a single warmizombie already done this?" Is it because there is any such data in the first place? Yes, I think that's it. I think the warmizombie congregation does not want anyone to realize that there isn't any Holy Grail Global Warming datafile out there. The warmizombies are not honest when they throw that term around "the data." There's a good reason they never specify what data. The reason is that it is completely imaginary data for whose existence they have profound religious faith and hope.

Surface Detail, I hate to burst your bubble. There is no Global Warming dataset that shows your religion to be true. I'm terribly sorry that this is the first you're hearing of this but don't worry, your cognitive dissonance is already at work making you .....



There, that's better. Was this another blank post by IBdaMann? Why does he do that?


.

jwoodward48 wrote:You're the only one who suspects the scientists of academic dishonesty.

jwoodward48 wrote:I trust that there's no HUMUNGUS HOACKS CONSPEERACEE. Why should I spend days digging through the raw data?

IBdaMann wrote:
There, that's better. Was this another blank post by IBdaMann? Why does he do that?

Surface Detail wrote: So why are you referring to yourself in the third person? Was that a bit of a slip? Are you part of a team effort?

IBdaMann wrote:

Surface Detail wrote: So why are you referring to yourself in the third person? Was that a bit of a slip? Are you part of a team effort?

It would be exceedingly delicious if you actually didn't get it.

jwoodward48 wrote:

IBdaMann wrote:

Into the Night wrote: You cannot use weighting. That violates the requirement of random selection. No data can substitute for missing data. Remember, the selection MUST be randN. Weighting the data effectively turns it into a randR.

Correct. "Weighting" is a polite word that warmizombies use to acknowledge that they have doctored/tweaked/cooked/fudged/skewed the raw data.

All unmodified raw data must be presented for scrutiny. Period.

All margins of error for all sensors/equipment involved must be presented. Period.

What is allowed, and encouraged, is to then also provide an analysis with weightings and explanations (also presented for scrutiny) if they make sense.

If I don't have all the unmodified raw data such that I can drop it all into a spreadsheet and run my own analyses then I have to ask what they are trying to hide. At what conclusions do they not want me to arrive such that they are providing me only modified data or worse, by only telling me the conclusions I am to believe without question?

.

Weighting means that close-together thermometers have less of an effect than further-apart thermometers, in order to give each region equal weighting. But then again, since "95% confidence level" apparently means "I'm really, really sure of it," who knows? The entire thing is a hoax. They're rewriting science at the moment! Run! Take this red pill and make a break for it!

jwoodward48 wrote:

IBdaMann wrote:

Surface Detail wrote: I just pointed out where you could download the raw data from.

Nope. You are simply weaseling again in your constant string of ongoing dishonesty. The story never changes.

1. I see your link and your promise of the Holy Grail of Global Warming data that shows everything you've been claiming. My search is over!

2. I go to your link and disappointingly I don't find the Holy Grail you promised. All I find is the following:

Mar 25 2008 00:00 10445 alaska-temperature-anomalies.txt
Mar 12 2008 00:00 3410 alaska-temperature-means.txt
Nov 10 2011 00:00 Directory anom
Sep 23 2016 13:03 Directory blended
Oct 17 2016 08:57 Directory daily
Dec 15 2009 00:00 Directory forts
Sep 11 2003 00:00 14953824 grid_gpcp_1979-2002.dat
Nov 30 2011 00:00 3886592 Lawrimore-ISTI-30Nov11.ppt
Feb 27 2013 00:00 Directory snow
Aug 22 2001 00:00 Directory v1
Oct 15 2016 09:00 Directory v2
Oct 17 2016 08:07 Directory v3
May 27 2016 09:03 Directory v4

3. I stand back and wait for you to pout like a fukcing baby and to scream "The data is THERE, you just have to search and hunt and scour and find it yourself! I'm not going to do your homework for you!"

4. I then tell you that you should do the legwork to go to all the places where you know the data to be, gather it, compile it yourself and put it into a consolidated Holy Grail of Global Warming datafile that I or anyone else can easily drop into a spreadsheet to perform analyses. If it's not important enough for you to do then it's not important enough for me either.

This brings us to the question of "Why hasn't a single warmizombie already done this?" Is it because there is any such data in the first place? Yes, I think that's it. I think the warmizombie congregation does not want anyone to realize that there isn't any Holy Grail Global Warming datafile out there. The warmizombies are not honest when they throw that term around "the data." There's a good reason they never specify what data. The reason is that it is completely imaginary data for whose existence they have profound religious faith and hope.

Surface Detail, I hate to burst your bubble. There is no Global Warming dataset that shows your religion to be true. I'm terribly sorry that this is the first you're hearing of this but don't worry, your cognitive dissonance is already at work making you .....



There, that's better. Was this another blank post by IBdaMann? Why does he do that?


.

You're the only one who suspects the scientists of academic dishonesty. I trust that there's no HUMUNGUS HOACKS CONSPEERACEE. Why should I spend days digging through the raw data?

Into the Night wrote:
This holy grail website does NOT have the raw data. It can't. No one ever put a thermometer in most places of the world. Possible temperature gradients vary too steeply and over too short a time for any instrumentation we now have.

Surface Detail wrote:

Into the Night wrote:
This holy grail website does NOT have the raw data. It can't. No one ever put a thermometer in most places of the world. Possible temperature gradients vary too steeply and over too short a time for any instrumentation we now have.

You're missing the point. We don't need to know the average temperature over a particular area, however you choose to define it. We just need a consistently measured average temperature since we're looking at temperature changes, not absolute temperatures.

Surface Detail wrote:
For example, say we have 1000 thermometers scattered across the globe, and the average temperature recorded by these thermometers rises from 16C to 17C over some period of time. Then we can be reasonably sure that the average temperature of the earth, whatever that may be, has also risen by about 1C.

Into the Night wrote:
No, it is you that is missing the point. You cannot claim a change unless you know the base of that change. That's absolute values.

Surface Detail wrote:

Into the Night wrote:
No, it is you that is missing the point. You cannot claim a change unless you know the base of that change. That's absolute values.

Of course you can measure a change without knowing the absolute values.

Let us define
T1 = Absolute average temperature at time 1 (unknown)
T2 = Absolute average temperature at time 2 (unknown)
M1 = Measured average temperature at time 1
M2 = Measured average temperature at time 2
D = Difference between measured and absolute average temperature values (unknown)

So M1 = T1 + D and M2 = T2 + D



Now, the change C in temperature between time 1 and time 2 is given by

C = T2 - T1

Bur T2 = M2 - D and T1 = M1 - D, so

C = (M2 - D) - (M1 - D) = M2 - M1

Hence we don't need to know the absolute values T1 and T2.

Into the Night wrote:

Surface Detail wrote:

Into the Night wrote:
No, it is you that is missing the point. You cannot claim a change unless you know the base of that change. That's absolute values.

Of course you can measure a change without knowing the absolute values.

Let us define
T1 = Absolute average temperature at time 1 (unknown)
T2 = Absolute average temperature at time 2 (unknown)
M1 = Measured average temperature at time 1
M2 = Measured average temperature at time 2
D = Difference between measured and absolute average temperature values (unknown)

So M1 = T1 + D and M2 = T2 + D



Now, the change C in temperature between time 1 and time 2 is given by

C = T2 - T1

Bur T2 = M2 - D and T1 = M1 - D, so

C = (M2 - D) - (M1 - D) = M2 - M1

Hence we don't need to know the absolute values T1 and T2.

You don't have a usable T1, T2, M1, or M2.

You are also attempting to declare Mx is somehow different than Tx.

Surface Detail wrote:Of course you can measure a change without knowing the absolute values.

Surface Detail wrote:

Into the Night wrote:

Surface Detail wrote:

Into the Night wrote:
No, it is you that is missing the point. You cannot claim a change unless you know the base of that change. That's absolute values.

Of course you can measure a change without knowing the absolute values.

Let us define
T1 = Absolute average temperature at time 1 (unknown)
T2 = Absolute average temperature at time 2 (unknown)
M1 = Measured average temperature at time 1
M2 = Measured average temperature at time 2
D = Difference between measured and absolute average temperature values (unknown)

So M1 = T1 + D and M2 = T2 + D



Now, the change C in temperature between time 1 and time 2 is given by

C = T2 - T1

Bur T2 = M2 - D and T1 = M1 - D, so

C = (M2 - D) - (M1 - D) = M2 - M1

Hence we don't need to know the absolute values T1 and T2.

You don't have a usable T1, T2, M1, or M2.

T1 and T2 are the unknown averages of the real temperatures; I see no reason why we shouldn't assign a label to these quantities. M1 and M2 are simply the averages of the measured temperatures, and I've just shown why we can use them.

You are also attempting to declare Mx is somehow different than Tx.

Well, yes, that's precisely because we can't accurately measure the real average temperatures. The measured temperatures will differ from the actual temperatures by some small error, which I've called D.

jwoodward48 wrote:
As an example, let us imagine a region. For ease of mathematics, let it be flat; this is not necessary, though, but lets you imagine the resulting functions as 3D shapes instead of 4D shapes. Much easier, right? This region can be represented by a domain A of the function T(x,y,t), which gives the temperature at a particular point in space and time.

Now, let's imagine an observatory at one point. It takes a single measurement at a single time. (Because we are looking at a single instant, time can be treated as a constant, and thus will not be discussed until later.) What is the margin of error of that measurement?

To answer this, let's imagine ourselves as omniscient observers who are capable of knowing the exact value of T(x,y) at any point. This means that we can find the average value of T (by taking the double integral and dividing by the area of A). Let us call this T_avg.

If, for 95% of the points in A, T(x,y) for that point is within m of T_avg, the 95% margin of error (henceforth referred to as the margin of error) is m. This means that if I take a single observation, it has a 95% of being within the corresponding confidence interval.

So now we know the value of m. But wait! There's more - this is only Region 1. We have nineteen more regions, each with a single observatory. How can we figure out the MoE now? Simply by taking the average of m_1, m_2, ... m_20. This makes intuitive sense, as the "effect of error" of a particular measurement is 20 times less when looking at an average of 20 numbers. If a particular measurement is 10K from the true average, than that measurement contributes 0.5K error to the average.

But the scientists in our scenario can't know the true m. They don't have T(x,y)! That's okay. All we need to know is the average margin of error for a region of a particular size. This is getting meta, but if we know with 95% confidence that the 95% MoE for a 10 mi2 region is less than 5K, then that is the MoE that we can use for a region of that size. This will be off sometimes - about 5% of the time. But that's okay! That's accounted for with the 95% CL.

But wait - what if the regions are sized differently? That's okay too! A region of larger size will have both a larger impact on the average and a larger MoE. As you can expect, sparser measurements have increasingly worse effects on the error, but this is quantifiable.

But even if the error in the absolute measurement is high, the error in the derivative is not necessarily high too. Let me demonstrate with an example from my young, sciency days. As opposed to my older sciency days.

======

jwoodward48 wrote:
When I was somewhat young, being a tiny 8th grader or so, my class made rockets and launched them. Nothing big, just paper rockets and an air-pressure launcher. We had previously made clinometers and used them to measure the height of various flagpoles and lightposts etc., and we used them again to measure the height of each rocket's trajectory.

Now, I, being one of scientific inclinations, wanted to test another way of measuring the height - timing how long the rocket is in the air. So my class did some trig on their calculators, and I did some algebra, and we all came up with our numbers. After doing some statistical analysis with Pierce's criterion, I weeded out the outliers from the class's results (some people can't even use a clinometer correctly, I tell you), and compared the two sets.

Mine were very incorrect. After some thought, I attributed this to my reaction time - the launch started with a "three, two, one," but the landing didn't, which would result in extra time being added to my numbers. To test this, I looked at each rocket's flight's height converted to seconds of flight - both my numbers and the class's. Indeed, mine was off by about 0.6 seconds every time.

So my method of observation was not useful for determining how high a rocket went. But was it useful for determining which one went higher? Both my data and the class's indicated the same winner, and a graph of my data and the class's looks very similar - my data is shifted upward, but otherwise identical.

Why is this useful? Because it is predictable error. It doesn't change over time.

======

jwoodward48 wrote:
In the temperature case, we can't know the error. We have nothing to calibrate with.

jwoodward48 wrote:
But we do know that some error is steady. If I stick my observatory beneath a cool overhang, that cool overhang will continue to be cooler than the rest of the region tomorrow, and the day after that.

jwoodward48 wrote:
This means that the margin of change for the derivative is less than the margin of error of the absolute measurements. This is why we use anomalies.

======
You cannot just choose a convenient population for your margin of error.
[quote]jwoodward48 wrote:
Also, as a reply to your "temperature changes greatly over distance" point and example, keep in mind that outliers do exist. This is fine. The average temperature change over distance is far less than what you observed.