Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Dave, if you can't follow the conversation, why not just sit quietly in the corner with your six-shooters and your comic books and wait till the grown-ups are done talking?
You couldn't carry a pencil for a real mathematician.. The purpose of a trend line is to estimate the 1st derivative of a data set. That is always a USEFUL estimate in that sense.. Even in the presence of data with high deviation or noise. No different than a selective filtering operation to accentuate different derivatives of the process.. Problem with poor math education is that they tell you WHAT TO DO -- but rarely explain what the tools are actually for.. Where your "education" failed you is that nobody is ATTEMPTING to find the underlying equation for the Earth temperature versus Time curve BECAUSE THERE ISN'T ONE.... So the "errors" in fit are MEANINGLESS. It is simply to establish the dtemp/dtime.... Go be a victim...
Or better yet -- go bug the market analysts on Wall Street for applying trend lines to high variance data.. Tell THEM they can't do that unless their p-values are appropriate..
@Flac
And yet, for allll your ranting, you've said nothing that changes the fact that the OP and the article that it highlights are irrelevant because the graph is meaningless. It is mesningless because the regression has no meaning.
Every statistic that is calculated using sum of squares has an associated p-value that defnes the significance of the statistic. And every coefficient has a conficdence imterval. Lacking those, the regression line is meaningless. And everything hinges on that. The line is insignificant. Watt's conclusions are insignificant. The OP is insignificant. And your opinions are insignificant.
If you cannot address the fundamentals of the science correctly, then nothing you have to present has any significance.
And it is obvious to everyone here that all you can do is whine.
Half my day is spent working with people that have a correct answer.. But they are working the wrong problem.. WHEN to apply certain tests and how to INTERPRET those tests is more important than mindlessly following the functions offered in Excel..
So --- If I run a 3 or 4 year low pass filter over the temp chart data to "smooth" the high frequency data --- Have I affected the 1st derivative (slope) estimate? (remember that error term of the regression is ASSUMED to have specific properties for the P-test)
After I filter and the variance is REDUCED -- fit that to the best trend line.. Is the P-Value BETTER or WORSE?
Actually -- the number of samples is far more important to the significance of the slope estimation in this case than goodness of the raw linear fit of the data.
Oh, eat shit, you pompous stuffed shirt. Your cult has been caught cooking the books more than once. Get it? AGW is based on lies. We prove it to you, and what's your response? Post more of their lies.Dave, if you can't follow the conversation, why not just sit quietly in the corner with your six-shooters and your comic books and wait till the grown-ups are done talking?
You couldn't carry a pencil for a real mathematician.. The purpose of a trend line is to estimate the 1st derivative of a data set. That is always a USEFUL estimate in that sense.. Even in the presence of data with high deviation or noise. No different than a selective filtering operation to accentuate different derivatives of the process.. Problem with poor math education is that they tell you WHAT TO DO -- but rarely explain what the tools are actually for.. Where your "education" failed you is that nobody is ATTEMPTING to find the underlying equation for the Earth temperature versus Time curve BECAUSE THERE ISN'T ONE.... So the "errors" in fit are MEANINGLESS. It is simply to establish the dtemp/dtime.... Go be a victim...
Or better yet -- go bug the market analysts on Wall Street for applying trend lines to high variance data.. Tell THEM they can't do that unless their p-values are appropriate..
@Flac
And yet, for allll your ranting, you've said nothing that changes the fact that the OP and the article that it highlights are irrelevant because the graph is meaningless. It is mesningless because the regression has no meaning.
Every statistic that is calculated using sum of squares has an associated p-value that defnes the significance of the statistic. And every coefficient has a conficdence imterval. Lacking those, the regression line is meaningless. And everything hinges on that. The line is insignificant. Watt's conclusions are insignificant. The OP is insignificant. And your opinions are insignificant.
If you cannot address the fundamentals of the science correctly, then nothing you have to present has any significance.
And it is obvious to everyone here that all you can do is whine.
Half my day is spent working with people that have a correct answer.. But they are working the wrong problem.. WHEN to apply certain tests and how to INTERPRET those tests is more important than mindlessly following the functions offered in Excel..
So --- If I run a 3 or 4 year low pass filter over the temp chart data to "smooth" the high frequency data --- Have I affected the 1st derivative (slope) estimate? (remember that error term of the regression is ASSUMED to have specific properties for the P-test)
After I filter and the variance is REDUCED -- fit that to the best trend line.. Is the P-Value BETTER or WORSE?
Actually -- the number of samples is far more important to the significance of the slope estimation in this case than goodness of the raw linear fit of the data.
A safe assumption is that the newer dataset is more accurate. That would be the paranoia-free position.
The difference between the two, as explained by the people who did the adjusting, is that such adjustments correct for errors and differences such as time-of day, location changes, analog to digital and general magic ;-),
Why go back and alter data from over 50 years ago? What is the reason for altering it?
Wow...
If you don't care whether its accurate, why do you care if it even exists?
itfitzme's biggest problem with trying to claim CO2 causation of temperature increase is the 'hidden factor fallacy (or fraud)'.
it's like a wife saying that the family is going into $250 debt per month because her husband spends $250 on booze every month. perfect correlation. but then the husband says it's her fault because she spends 250 bucks a month on cigarettes. again perfect correlation. it depends on who defines the factors to be analyzed, and even whether all the factors are known. CO2 doesnt work so well in explaining the MWP or the LIA. but for the 80's and 90's it was perfect. now, not so much.
So, back to walleyes question if a lowpass filter has an associated probability and a confidence interval.
The answer can be had simply by asking of a ramping sinusoid can be sampled such that the result is a decreasing line.
Of walleyes is half as smart as he think he is, he can answer that easily. He will surely squirm away as usual.
The answer is absolutely. Every descrete time sampling filtere has a probability of yielding an output that isn't representative of the original signal. Digital communications requires statistics and probability as part of signal analysis and filter design. Every communications engineer studies statistics.
A sample of yeary temp is a descrete time sample.
That just proves that, yes, Watts graph is meaningless because no confidence interval and p- value is given.
The problem, which I'll get to later, is even worse because it is obvious by inspection thatr the confidence interval is so large thay the linear regression is meaningless.
Being an asshole, waleyed, doesn't make you any less wrong. I get that all that adrenaline makes you feel like you are. What is does do is make you both an asshole and stupid. There's a common phrase for that ..... what is it? You are a what?
Do you really need me to show you a sampled ramped sinusoid that gives a decreasing line, or can you manage that on your own?
I'd like to see you stop being intentionally wrong. I doubt that willl happen.
Well, back to work.
A safe assumption is that the newer dataset is more accurate. That would be the paranoia-free position.
The difference between the two, as explained by the people who did the adjusting, is that such adjustments correct for errors and differences such as time-of day, location changes, analog to digital and general magic ;-),
Why go back and alter data from over 50 years ago? What is the reason for altering it?
Wow...
If you don't care whether its accurate, why do you care if it even exists?
Why go back and alter data from over 50 years ago? What is the reason for altering it?
Wow...
If you don't care whether its accurate, why do you care if it even exists?
adjustments circa 2000
adjustments circa 2013
what is interesting to me is what doesnt seem to be on the graphs. no marked spike in corrections when a wholesale conversion to digital thermometers and 'morning readings'. at least in the 2000 graph the adjustments levelled off a bit after 1990. but that plateau disappeared in the 2013 edition.
the early work into TOBS used coarse decadal estimate figures for time of temperature readings. those coarse estimates didnt show up then, and they dont show up now on individual stations that should only have one TOBS adjustment per change in observation time.
@Flac
And yet, for allll your ranting, you've said nothing that changes the fact that the OP and the article that it highlights are irrelevant because the graph is meaningless. It is mesningless because the regression has no meaning.
Every statistic that is calculated using sum of squares has an associated p-value that defnes the significance of the statistic. And every coefficient has a conficdence imterval. Lacking those, the regression line is meaningless. And everything hinges on that. The line is insignificant. Watt's conclusions are insignificant. The OP is insignificant. And your opinions are insignificant.
If you cannot address the fundamentals of the science correctly, then nothing you have to present has any significance.
And it is obvious to everyone here that all you can do is whine.
Half my day is spent working with people that have a correct answer.. But they are working the wrong problem.. WHEN to apply certain tests and how to INTERPRET those tests is more important than mindlessly following the functions offered in Excel..
So --- If I run a 3 or 4 year low pass filter over the temp chart data to "smooth" the high frequency data --- Have I affected the 1st derivative (slope) estimate? (remember that error term of the regression is ASSUMED to have specific properties for the P-test)
After I filter and the variance is REDUCED -- fit that to the best trend line.. Is the P-Value BETTER or WORSE?
Actually -- the number of samples is far more important to the significance of the slope estimation in this case than goodness of the raw linear fit of the data.
You just add more bs on top of bs.
The trend line is a best estimate. A best estimate has a confidence interval. It also has a p-value, the probability of getting the same result from random chance. No published confidence interval and no p-value, no meaning.
The statistics are affected by a number of factors, sample size and variance are two. The slope is as well.
You csn prove that the trend line is precise, just by coming up with the p-value and confidence interval. It's that simple. No need for bullshit about low pass filters. Just do the stats correctly.
So, back to walleyes question if a lowpass filter has an associated probability and a confidence interval.
The answer can be had simply by asking of a ramping sinusoid can be sampled such that the result is a decreasing line.
Of walleyes is half as smart as he think he is, he can answer that easily. He will surely squirm away as usual.
The answer is absolutely. Every descrete time sampling filtere has a probability of yielding an output that isn't representative of the original signal. Digital communications requires statistics and probability as part of signal analysis and filter design. Every communications engineer studies statistics.
A sample of yeary temp is a descrete time sample.
That just proves that, yes, Watts graph is meaningless because no confidence interval and p- value is given.
The problem, which I'll get to later, is even worse because it is obvious by inspection thatr the confidence interval is so large thay the linear regression is meaningless.
Being an asshole, waleyed, doesn't make you any less wrong. I get that all that adrenaline makes you feel like you are. What is does do is make you both an asshole and stupid. There's a common phrase for that ..... what is it? You are a what?
Do you really need me to show you a sampled ramped sinusoid that gives a decreasing line, or can you manage that on your own?
I'd like to see you stop being intentionally wrong. I doubt that willl happen.
Well, back to work.
Are you on drugs? This is one of the most incoherent posts I've seen in a while. You OK?
You couldn't carry a pencil for a real mathematician.. The purpose of a trend line is to estimate the 1st derivative of a data set. That is always a USEFUL estimate in that sense.. Even in the presence of data with high deviation or noise. No different than a selective filtering operation to accentuate different derivatives of the process.. Problem with poor math education is that they tell you WHAT TO DO -- but rarely explain what the tools are actually for.. Where your "education" failed you is that nobody is ATTEMPTING to find the underlying equation for the Earth temperature versus Time curve BECAUSE THERE ISN'T ONE.... So the "errors" in fit are MEANINGLESS. It is simply to establish the dtemp/dtime.... Go be a victim...
Or better yet -- go bug the market analysts on Wall Street for applying trend lines to high variance data.. Tell THEM they can't do that unless their p-values are appropriate..
@Flac
And yet, for allll your ranting, you've said nothing that changes the fact that the OP and the article that it highlights are irrelevant because the graph is meaningless. It is mesningless because the regression has no meaning.
Every statistic that is calculated using sum of squares has an associated p-value that defnes the significance of the statistic. And every coefficient has a conficdence imterval. Lacking those, the regression line is meaningless. And everything hinges on that. The line is insignificant. Watt's conclusions are insignificant. The OP is insignificant. And your opinions are insignificant.
If you cannot address the fundamentals of the science correctly, then nothing you have to present has any significance.
And it is obvious to everyone here that all you can do is whine.
Half my day is spent working with people that have a correct answer.. But they are working the wrong problem.. WHEN to apply certain tests and how to INTERPRET those tests is more important than mindlessly following the functions offered in Excel..
So --- If I run a 3 or 4 year low pass filter over the temp chart data to "smooth" the high frequency data --- Have I affected the 1st derivative (slope) estimate? (remember that error term of the regression is ASSUMED to have specific properties for the P-test)
After I filter and the variance is REDUCED -- fit that to the best trend line.. Is the P-Value BETTER or WORSE?
Actually -- the number of samples is far more important to the significance of the slope estimation in this case than goodness of the raw linear fit of the data.
@Flac
And yet, for allll your ranting, you've said nothing that changes the fact that the OP and the article that it highlights are irrelevant because the graph is meaningless. It is mesningless because the regression has no meaning.
Every statistic that is calculated using sum of squares has an associated p-value that defnes the significance of the statistic. And every coefficient has a conficdence imterval. Lacking those, the regression line is meaningless. And everything hinges on that. The line is insignificant. Watt's conclusions are insignificant. The OP is insignificant. And your opinions are insignificant.
If you cannot address the fundamentals of the science correctly, then nothing you have to present has any significance.
And it is obvious to everyone here that all you can do is whine.
Half my day is spent working with people that have a correct answer.. But they are working the wrong problem.. WHEN to apply certain tests and how to INTERPRET those tests is more important than mindlessly following the functions offered in Excel..
So --- If I run a 3 or 4 year low pass filter over the temp chart data to "smooth" the high frequency data --- Have I affected the 1st derivative (slope) estimate? (remember that error term of the regression is ASSUMED to have specific properties for the P-test)
After I filter and the variance is REDUCED -- fit that to the best trend line.. Is the P-Value BETTER or WORSE?
Actually -- the number of samples is far more important to the significance of the slope estimation in this case than goodness of the raw linear fit of the data.
Half my day is spent working with people that have a correct answer.. But they are working the wrong problem..
So everyone at work think you're an idiot too.
@Flac
And yet, for allll your ranting, you've said nothing that changes the fact that the OP and the article that it highlights are irrelevant because the graph is meaningless. It is mesningless because the regression has no meaning.
Every statistic that is calculated using sum of squares has an associated p-value that defnes the significance of the statistic. And every coefficient has a conficdence imterval. Lacking those, the regression line is meaningless. And everything hinges on that. The line is insignificant. Watt's conclusions are insignificant. The OP is insignificant. And your opinions are insignificant.
If you cannot address the fundamentals of the science correctly, then nothing you have to present has any significance.
And it is obvious to everyone here that all you can do is whine.
Half my day is spent working with people that have a correct answer.. But they are working the wrong problem.. WHEN to apply certain tests and how to INTERPRET those tests is more important than mindlessly following the functions offered in Excel..
So --- If I run a 3 or 4 year low pass filter over the temp chart data to "smooth" the high frequency data --- Have I affected the 1st derivative (slope) estimate? (remember that error term of the regression is ASSUMED to have specific properties for the P-test)
After I filter and the variance is REDUCED -- fit that to the best trend line.. Is the P-Value BETTER or WORSE?
Actually -- the number of samples is far more important to the significance of the slope estimation in this case than goodness of the raw linear fit of the data.
Half my day is spent working with people that have a correct answer.. But they are working the wrong problem..
So everyone at work think you're an idiot too.