The Holocene Deniers

[Contracelsus]

If belief in global warming causes one to claim that Belle Plaine and Toledo have identical temperatures when the chart looks like that in post 192, then I would say it is pathalogical

I don't believe I claimed the two stations were perfectly identical. If I did I was in error. However I did point out that they only differ by 1 degree F (median value) and that was from a rather narrow distribution with long tails.

I hope I have been quite clear. That is why I have been posting all of my data and calculations.

(And please do keep in mind that I have been downloading the data and processing it here. I am doing quite a bit of work here. I would greatly appreciate it if my efforts were not treated as if I were merely doing knee-jerk reaction and gut-feel stuff. Thanks)

 

[Contracelsus]

I was looking back over my posts and I saw this:

So what do I believe? If I'm doing the stats right here then it looks to me like the counts you provided don't give me enough reason to say there is a directional bias at 20deg, 10deg or even down at 5deg.

But like I said, I could be doing this wrong. I am not a scientist. I don't do this stuff for a living.

In the two Iowa stations (Belle Plain and Toledo) I don't see the bias in the histogram of differences between readings. (And I have attempted to make sure they line up and I've subtracted only days that are the same)

My fascination is that, even when we think we see a bias sometimes that isn't a real bias. That's why statistics are fascinating things.

I was unclear in this post. I think the 1 degree median difference is real. That is a bias. But as I also said in the same post I don't see bias at higher temperature differences (5 degrees, 10 degrees and 20 degrees). In the present case there is a bias but only of 1 degree F. And there is virtually no skew to the data.

I simply don't see a 1 degree difference as being overly problematic, especially if the absolute temperature is not what is used by climatologists, but rather the change in temperature.

 

[Contracelsus]

For those who don't think there is a bias problem I want to post two sets of two towns. Coldwater Kansas and Ashland Kansas.

I downloaded this data as well. Obviously I already posted the difference histogram. But I also took the daily data and painstakingly went through as usual and found all places where there weren't matching data and eliminated them.

Then I plotted them with one station on the X axis and one station on the Y axis.

This was pretty fascinating. Sure there's an offset. We already determined that the data was not perfectly without difference, but in the present debate we are more interested in the trends of the temperature.

If there was a problem with two stations then they shouldn't "track" each other. In other words the slope shouldn't be 1. But in this case the slope is 0.948 which is pretty good evidence that the two stations track each other.

With a little time I might be able to get a statistic on the robustness of that slope, but it looks pretty good.

(Sorry I've got second shift today so I got some time this morning to do this. And besides this is a lot more fun that what I have to do the rest of the day!)

 

[grmorton]

Mr. Morton, I have downloaded all the data we've discussed. There is simply no other way to generate these histograms. Believe me I have put in a goodly amount of work downloading the data and then finding the places where the days don't line up and then loading it into a statistics package and running the tests.

There simply is no other way to generate these histograms.

You were using the counts I provided as far as I could tell. But you are correct it takes lots of work to look at the raw data. They don't make it easy.

Actually you'll note I didn't say there was absolutely no bias.

I am glad that you too can see that Belle Plaine's temperature is almost always higher than that of Toledo's just 18 miles away.

My data found 1 degree difference between the station as the median difference. However, I do not find 1 degree difference to be startling.

That is the average. You ignore the 25% of the days where the temperature difference is greater than 3 degrees. Indeed, the average of the absolute value of the temperature difference is 2.7 deg. That is over a record since 1948. By taking out the sign of the difference one can see the magnitude of the average temperature difference. It seems that on average there is a temperature gradient equal to a moderate cold front.

You will note, that I was testing your assessment that there were biases at higher degree differences which I failed to find.

And as far as I can tell failed to actually mention. But your methodology was flawed. All one has to do to see a big bias is to look at the plots of the annual average temperature--which I posted.

I am moving on to other data sets. I think we have discussed this one enough.

Tonight I am going to post a picture for two UK towns, Ringway and Shawbury. One can see that when I subtract the two temperature trends for these two towns merely 51 miles apart, since 1988 Ringway has experienced more 'global warming' than Shawbury. The warming in Ringway doesn't seem to extend to Shawbury but CO2 should have the same effect in both towns unless people think that CO2 likes the pubs in Ringway better than those in Shawbury.

I also want to post a picture from China weather stations, which are included in the global warming calculation and are downloadable from the Beijing weather service. They only give yearly averages so it is difficult to see what the problem is but when I plot two nearby stations (based on the lat/long data), I find some really confidence building examples of the competence of meteorologists to measure the weather. That is the second picture.

The UK picture was converted to deg F for US readers, the Chinese data was not, so the temperature difference in Fahrenheit between the two Chinese stations is almost double the number one gets by subtracting the two curves.

The Chinese data is near Guangzhou, a city I have been to many many times. It is a beautiful city in subtropical China. The difference in temperature between these two towns is as much as 10 deg C, which is 28 deg F. And this is a yearly average temperature, not a daily temperature. That means that the temperature for station 79 for 2001 must have been 10 degrees C hotter on average every single day (18 deg F. Of course, in 1968 it was 4 deg C or approx 7 deg F colder.

Such lunatic variances in temperature make it insane to think that we can use the meteorological temperature data to determine a trend.

Does anyone here want to say that the Chinese record is good?

The third picture is the data from these two stations

By the way Contracelsus, you still haven't addressed the physics issue. Talking about data distribution won't cut it. Do you think it is reasonable for a strong cold front temperature gradient to exist between these two towns for 25% of the days over the past 62 years?? Please answer the question. YOu didn't answer the physics issue.

 

[grmorton]

I downloaded this data as well. Obviously I already posted the difference histogram. But I also took the daily data and painstakingly went through as usual and found all places where there weren't matching data and eliminated them.

Then I plotted them with one station on the X axis and one station on the Y axis.

This was pretty fascinating. Sure there's an offset. We already determined that the data was not perfectly without difference, but in the present debate we are more interested in the trends of the temperature.

If there was a problem with two stations then they shouldn't "track" each other. In other words the slope shouldn't be 1. But in this case the slope is 0.948 which is pretty good evidence that the two stations track each other.

With a little time I might be able to get a statistic on the robustness of that slope, but it looks pretty good.

(Sorry I've got second shift today so I got some time this morning to do this. And besides this is a lot more fun that what I have to do the rest of the day!)

Interesting plot you have there. Frankly I think it supports my contention that we are not measuring the data temperature with any level of accuracy. The width of your cloud is about 40 degrees F wide. First off, I think you should check to see that you have things aligned correctly and secondly that you don't have the missing data contributing to the above chart. I expect a 20 deg F width to the graph. My suspicion is that you make the data actually look worse than what it should. Look at the town represented by the x axis. When it is 40 degrees, the other town 18 miles away can be anywhere from 20 degrees to 60 degrees. That means the error bar in thermometer measurement assuming this is 3 SD (I didn't calculate that) is +/- something like 6 degrees.

If you look at the 40 deg F temperature on the town on the Y axis. When that town is 40 deg F. the other town can be anywhere from 14-64 deg F.

If you think a temperature system which measures the temperature only within the accuracy of +/-6 degrees is capable of measuring a global temperature rise of 1 deg F you clearly have some work to do in statistics.

Frankly, I think your plot supports beautifully my position that the temperature record is crap for the purposes of global warming measurments

Secondly, No one said, certainly not me, that the two stations don't move in general in sync. That isn't a claim you can find me making.

I am going to ask again, do you think that the temperature gradient appropriate to a strong cold front existed between these two towns for 25% of the days. If you answer no, then the data is crap. If you answer yes, you have said that there exists a novel meteorological phenomenon.

I forgot to post this earlier. Some might want to look at the efforts to cap carbon emissions in Australia. It mentions Plimer whome Gracchus seemed not to think was really a geologist.
http://www.realclearpolitics.com/ar...art_the_great_global_warming_scare_97148.html

 

[Contracelsus]

You were using the counts I provided as far as I could tell.

When I did the chi squared tests. When I did the histograms I needed to download all the data. (That's why the histograms were as large as they were.) MY first post in this thread used downloaded data, complete sets. So I've been downloading the full sets from the start here.

But you are correct it takes lots of work to look at the raw data. They don't make it easy.

Agreed.

That is the average. You ignore the 25% of the days where the temperature difference is greater than 3 degrees. Indeed, the average of the absolute value of the temperature difference is 2.7 deg. That is over a record since 1948. By taking out the sign of the difference one can see the magnitude of the average temperature difference. It seems that on average there is a temperature gradient equal to a moderate cold front.

If I weigh myself on two bathroom scales every day. One is offset from the other by 1 lb I can still track the increase or decrease of my weight.

And as far as I can tell failed to actually mention.

Actually in the chi square discussion I was pretty explicit about that. In fact that was the only thing I could test for with that data. (http://www.christianforums.com/t7394133-17/#post52740154)

But your methodology was flawed.

Then please show me how the chi square calculation was in error. For 20 deg, 10 deg and 5 deg the chi square test resulted in insignificant p values which means one cannot reject the null hypothesis.

All you need do is show me in the calculation where I miscalculated.

All one has to do to see a big bias is to look at the plots of the annual average temperature--which I posted.

No, the "big bias" was a 1 deg median difference. Sure there's larger differences on different days but the median of the data, the median of the difference over 60 years was 1 degree.

Your graph showed a difference but my histogram clarified how large that difference came out to in the overall data set.

I am moving on to other data sets. I think we have discussed this one enough.

Excellent idea.

Tonight I am going to post a picture for two UK towns, Ringway and Shawbury.

I will have to have the raw data. After all that is why I found this discussion interesting. No filters. Just raw data.

One can see that when I subtract the two temperature trends for these two towns merely 51 miles apart, since 1988 Ringway has experienced more 'global warming' than Shawbury. The warming in Ringway doesn't seem to extend to Shawbury but CO2 should have the same effect in both towns unless people think that CO2 likes the pubs in Ringway better than those in Shawbury.

I don't have the raw data, but I will grant there is a trend in the differences there. Of course the average difference looks at first glance as if it will still work out to be only a couple of degrees.

At least in this case the trend does show that it would be difficult to consider these two as reasonably good replicates.

The Chinese data is near Guangzhou, a city I have been to many many times. It is a beautiful city in subtropical China.

I liked Guangzhou as well. I only got to stay a couple days. Had a great adventure on an afternoon trip to Hong Kong by rail when I found out that China, even though it now controls Hong Kong likes to require a separate entry visa to get back in. I got a nice night in a nice hotel in Hong Kong.

Such lunatic variances in temperature make it insane to think that we can use the meteorological temperature data to determine a trend.

But again, the key that keeps getting ignored is that the temperature data is averaged over large areas so any consistent temperature differences (as in the prior discussion, where the difference median was 1 deg F) aren't as problematic. The trends in global warming are not based on single stations. Nor should they be. Of course there's going to be errors. But if we must address each station individually then it is important that we use a statistically robust analysis.

By the way Contracelsus, you still haven't addressed the physics issue.

That's OK, you haven't directly addressed any of the errors you seem to see in my statistics. A calculation would be helpful. You also haven't addressed my question about what would be your preferred type of data set. One without any errors whatsoever? Would you trust it if there was absolutely no error?

Talking about data distribution won't cut it.

I thought you were the guy who said:

Finally someone who actually likes to analyze data

If your going to pick apart the data on a station-by-station basis I think it quite reasonable to discuss the data at this level. That is precisely why statistics is powerful.

Do you think it is reasonable for a strong cold front temperature gradient to exist between these two towns for 25% of the days

I thought I had addressed this. Mea culpa if I hadn't. First of all I am unsure how to differentiate between a legitimate cold front between two town and actual plain ol' error in the measurement. Who knows? Maybe one day Clyde the temperature reader in 1953 was too cold to get the temp right so he wrote down a number that was 5 degrees off of what it was really?

I think this is the importance of looking at the statistics. Errors are found everywhere. If there's a system that produces error-free data then I am afraid it would raise more eyebrows and probably be less trusted than normal things here in the physical world.

over the past 62 years?? Please answer the question. YOu didn't answer the physics issue.

Well now I have. You are drawing conclusions around the error in the data and asking if I believe it to be perfectly real. Clearly I do not believe it is likely that on Aug 13, 19XX there was a 40 degree difference between the stations. However I am fully aware of error in any measurement system.

Now I have addressed your physics question. If you could please return the favor and address the reasons why the statistics I have calculated are in error. I have provided numerous citations as to which statistical tests I am using, I can also provide you with author/title and publisher of the stats books I have here at home if you like. We can work through the problems together.

And please do be aware that part of the fun for me is to analyze the data, not just look at the pictures. If I can learn more statistics then I consider this to be a very fruitful conversation.

 

[Contracelsus]

Interesting plot you have there. Frankly I think it supports my contention that we are not measuring the data temperature with any level of accuracy. The width of your cloud is about 40 degrees F wide.

That's where the histogram of differences comes in handy! Take a look back at the histogram in the earlier post.

My suspicion is that you make the data actually look

I would gladly take a look at something more than your suspicion.

If you think a temperature system which measures the temperature only within the accuracy of +/-6 degrees is capable of measuring a global temperature rise of 1 deg F you clearly have some work to do in statistics.

I think I have more than amply addressed the point here:
http://www.christianforums.com/t7394133-20/#post52756778

If you would like I can run new numbers for you.

If I make up two data sets one with mean = 20 deg and one with mean = 21 degrees and both have standard deviation of 6 here's what I get:

Two Sample t-test

data: data1 and data2
t = -16.1949, df = 39998, p-value < 2.2e-16
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-1.089676 -0.854391
sample estimates:
mean of x mean of y
20.02143 20.99346

So again it looks like I can easily differentiate two distributions whose mean only differs by one degree and whose standard deviation is 6 degrees.

That's what that really small p-value is. "If p is low the null must go" as they say.

(I should point out that in this example as in the earlier one the two distributions are ideally normal, that's one of the requirements of the t-test, so a t-test comparing two more "peaked" distributions might alter the outcome of the t-test--it is technically a violation of one of the assumptions of the t-test, but in dealing with climate data I believe they are usually looking at large averaged data sets whose distributions I am unaware of. In the case of the peaked distributions of the individual stations it is likely that the standard deviation may be somewhat inflated and end up lowering the t-value a bit. But I'm not terribly sure if I've got that right.

Here's something I found about the robustness of the t-test in cases where the assumption of normality is violated:

Unless the sample size is small (less than 10), light-tailedness or heavy-tailedness will have little effect on the t statistic. ... Heavy-tailedness will tend to increase the chance that an incorrectly large P value will be reported (i.e., that the null hypothesis will not be rejected when it is in fact false, making the test conservative.

From: (Does your data violate one-sample t test assumptions?)

(So at least it would err on the conservative side and fail to see differences when differences exist).

Secondly, No one said, certainly not me, that the two stations don't move in general in sync. That isn't a claim you can find me making.

Excellent because that is the important bit from what I've read in the climate stuff.

 

[grmorton]

That's where the histogram of differences comes in handy! Take a look back at the histogram in the earlier post.

I would gladly take a look at something more than your suspicion.

I think I have more than amply addressed the point here:
http://www.christianforums.com/t7394133-20/#post52756778

No, that post doesn't do it.


Are you ever going to actually address the issue of the temperature gradient or are you going to keep ducking that issue?

 

[grmorton]

I want to go back to an issue raised by Thistlethorn. In a previous post I put up the degree-days above zero C for Tura Russia. To refresh everyone a degree-day above zero is the number of degrees above zero times the number of days that achieve this level. Then I summed them up for each year. If Siberia is warming and melting, as the hysteriacs claim, then we should see more days above zero and higher degrees. The multiplication of the degrees times the day should make the degree-days chart shoot up rapidly in a warming environment.

Anyway, when I posted Tura showing cooling (fewer degree-days), Thistlethorn dismissed it as being a local affair Tura is at 64N 100E. I will put up Ostrov Kotel' which is at 76 N 37 E way on the east side of Russia and Isit Russia at 60 N 125 E. that is almost 100 degrees of longitude separation. Even at these latitudes this is a long way away. Note that the degree days are decreasing, not increasing as global warmists say.

So, why is Siberia always shown as warming? Because the night time temperatures have warmed, but if it is going to MELT, it must have more time ABOVE zero C and the actual data says that the time spent above zero over the past century has declined.

So, the alarmists use the average temperature to make it appear as if Siberia is melting, when in fact, it is doing no such thing as it is spending more time below zero.

 

[Thistlethorn]

I want to go back to an issue raised by Thistlethorn. In a previous post I put up the degree-days above zero C for Tura Russia. To refresh everyone a degree-day above zero is the number of degrees above zero times the number of days that achieve this level. Then I summed them up for each year. If Siberia is warming and melting, as the hysteriacs claim, then we should see more days above zero and higher degrees. The multiplication of the degrees times the day should make the degree-days chart shoot up rapidly in a warming environment.

Anyway, when I posted Tura showing cooling (fewer degree-days), Thistlethorn dismissed it as being a local affair Tura is at 64N 100E. I will put up Ostrov Kotel' which is at 76 N 37 E way on the east side of Russia and Isit Russia at 60 N 125 E. that is almost 100 degrees of longitude separation. Even at these latitudes this is a long way away. Note that the degree days are decreasing, not increasing as global warmists say.

So, why is Siberia always shown as warming? Because the night time temperatures have warmed, but if it is going to MELT, it must have more time ABOVE zero C and the actual data says that the time spent above zero over the past century has declined.

So, the alarmists use the average temperature to make it appear as if Siberia is melting, when in fact, it is doing no such thing as it is spending more time below zero.

No, they use the average temperature because that what's interesting. I've already deconstructed your irrelevant degree-day data in a previous post. Since then, all you have done is continuing arguing about red herrings, and obfuscating what climatologists are really saying.

I wish you a nice life. You obviously have no interest in discussing the real issues here, but would rather keep throwing red herrings into the discussion. You are a stereotypical denier, with obvious political reasons for denying global warming and with enough science education to keep throwing irrelevant data into the discussion, but not enough education to understand what the experts are really talking about. Perhaps you can tell your children that you did your part to keep poisoning the earth.

 

[Contracelsus]

No, that post doesn't do it.


Are you ever going to actually address the issue of the temperature gradient or are you going to keep ducking that issue?

Please, Mr. Morton, I have addressed that issue now a couple of times. I am unsure how my explanation has failed to resonate with you but I will reiterate it:

Data has errors. Sometimes those errors are human, sometimes those errors are instrumental. That's why data has distributions.

I do not believe that every time there was a 20degree difference between the two stations that there was necessarily a 20 degree difference in actual temperature. Hence the use of the word error.

I am sure there were times when there was a measurable real difference in temeperature between the two stations that was derived from the average temperature for that day. Sometimes a cold front might have moved through.

I do not know how much more explicit I can be on this issue. This is the very nature of error. It sometimes just is. In cases of actual error there is no real data there, it is an error. In other cases there probably was a cold-front I don't know.

Now I have addressed the issue. If it is that important to know how much of the 25% of the error is actual error and how much was actual cold-front or weird weather I would recommend going back through 60 years worth of daily weather reports (that's 21,900 days worth of weather data) and parsing out which is which. I would suspect at that time that you will find that sometimes (especially with the larger differences) it was just error, pure error.

 

[Contracelsus]

Are you ever going to actually address the issue of the temperature gradient or are you going to keep ducking that issue?

Mr. Morton, can I ask why you doggedly insist everyone address every single point you make in detail, while you seem to go right by others' points?

For instance you have now claimed on a couple of occasions the following:

If you think a temperature system which measures the temperature only within the accuracy of +/-6 degrees is capable of measuring a global temperature rise of 1 deg F you clearly have some work to do in statistics.

You just illustrated the problem. If the noise is greater than the signal, it gets very hard to see the signal. the signal in this case is 1.1 deg F of global warming, but the noise is 5 degrees (using your example above). I will never be convinced that it makes sense to say that the globe has warmed by 1.1 deg +/- 5 degrees.

I addressed that point with examples and calculations now twice:

Here:http://www.christianforums.com/t7394133-20/#post52756778

and

Here: http://www.christianforums.com/t7394133-21/#post52761699

Could you please address the points I have raised to counter your claims? It would be only fair considering that you insist everyone address your points explicitly.

I would very much enjoy understanding the mathematics underlying your claim and how it would seem to differ from the examples I gave in which differences that were up to 6 times less than the standard deviation of "model data" sets were still detectable.

I will grant I may be missing something here but just telling me I'm missing it won't cut it. I must have the mathematical justification for the failure. That is the rigor which I thought this debate was all about.

 

[grmorton]

No, they use the average temperature because that what's interesting. I've already deconstructed your irrelevant degree-day data in a previous post. Since then, all you have done is continuing arguing about red herrings, and obfuscating what climatologists are really saying.

We will disagree on this. The issue in Siberia is whether or not the permafrost is melting. Yet, amazingly, there is, with each passing year, less opportunity for it to melt. The Degree-days are fewer over all the cities in Siberia save 3. I looked at 31 different stations. So, by use of the average, in which the monthly minimum temperature warms but the monthly maximum temperature cools the average goes up but the chance for melting goes down. Yet they claim that all the permafrost is about to melt.

the same thing is happening at Barrow Alaska. See below.

I wish you a nice life. You obviously have no interest in discussing the real issues here, but would rather keep throwing red herrings into the discussion. You are a stereotypical denier, with obvious political reasons for denying global warming and with enough science education to keep throwing irrelevant data into the discussion, but not enough education to understand what the experts are really talking about. Perhaps you can tell your children that you did your part to keep poisoning the earth.

Thank you I have a nice life. I think you mean I have no interest in looking at a chart that has a +/- 20 deg F temperature spread and agreeing with you that we are measuring the temperature marvelously. In that you would be correct.

But before you go, would you please answer the question you keep ducking? A strong cold front has a temperature gradient of 0.2 deg F per mile. Do you think it is reasonable for 25% of the days to have a greater temperature gradient than a strong cold front?

 

[grmorton]

Please, Mr. Morton, I have addressed that issue now a couple of times. I am unsure how my explanation has failed to resonate with you but I will reiterate it:

Data has errors. Sometimes those errors are human, sometimes those errors are instrumental. That's why data has distributions.

Because you are not talking about physics. You are talking about distributions. Cold fronts have strong temperature gradients. If your answer doesn't include the word 'gradient' then you can't possibly be answering the question I asked.

I do not believe that every time there was a 20degree difference between the two stations that there was necessarily a 20 degree difference in actual temperature. Hence the use of the word error.

Yes, a 20 degree difference in temperature over 18 miles indicates a HUGE error in the thermometer record. One or the other of the two towns is WRONG in its temperature. When these errors are so large that they make no physical sense, then one has to ask the question whether or not one is actually achieving the goal of accurately measuring the temperature.

I am under the impression that the goal is to accurately measure the temperature is it not? If so, why doesn't someone actually try to figure out why 25% of the days have impossibly large temperature differences?

I am sure there were times when there was a measurable real difference in temeperature between the two stations that was derived from the average temperature for that day. Sometimes a cold front might have moved through.

No doubt, cold fronts goe through these two towns. But on 1 out of 4 days???? Are you serious?

I do not know how much more explicit I can be on this issue. This is the very nature of error. It sometimes just is. In cases of actual error there is no real data there, it is an error. In other cases there probably was a cold-front I don't know.

Well if the error is greater than the purported warming you are in the silly position of saying that the world has warmed .84 +/- 4.07 degrees. Meaning, that it might have cooled by 3 degrees or warmed by 5. With errors like this, no one can know if the two cities have experienced global warming. In other words, the thermometer record is crap.

4.07 happens to be the standard deviation of the delta temperature between these two towns. On the vast vast majority of days, the average temperatures should be almost the same, not 4 degrees different or evern 20 degrees different.

Now I have addressed the issue. If it is that important to know how much of the 25% of the error is actual error and how much was actual cold-front or weird weather I would recommend going back through 60 years worth of daily weather reports (that's 21,900 days worth of weather data) and parsing out which is which. I would suspect at that time that you will find that sometimes (especially with the larger differences) it was just error, pure error.

Remember to get that much temperature difference, the front must stall out IN BETWEEN the two towns for the whole day. That isn't going to happen more than about once every 10 years.

 

[grmorton]

Mr. Morton, can I ask why you doggedly insist everyone address every single point you make in detail, while you seem to go right by others' points?

Because I am dogged; because I won't let you ignore important points

For instance you have now claimed on a couple of occasions the following:

http://www.christianforums.com/t7394133-21/#post52761699

Could you please address the points I have raised to counter your claims? It would be only fair considering that you insist everyone address your points explicitly.

Given that you think this kind of error is ok, and given that you keep talking about distributions rather than in terms of physics and meteorlogy, I don't find that your 'counter points' are really even relevant to the question I am asking.

I would very much enjoy understanding the mathematics underlying your claim and how it would seem to differ from the examples I gave in which differences that were up to 6 times less than the standard deviation of "model data" sets were still detectable.

Temperature differences are detectable by the mere fact of subtraction. That is an interesting mathematical invention from about 3500 BC. I can even detect differences at the 20,000ths standard deviation by mere subtraction. If I have a data set with a std. dev. of 1, if two values are different by 6 standard deviations, then subtraction will give an answer of 6. If two of the points are separated by 20,000 sd, then the subtraction will yield 20,000. It isn't that hard Contracelsus.

The hard thing comes from measuring a 1 degree change when the noise is 6x that

I will grant I may be missing something here but just telling me I'm missing it won't cut it. I must have the mathematical justification for the failure. That is the rigor which I thought this debate was all about.

I just told you what you were missing

 

[Contracelsus]

Because you are not talking about physics. You are talking about distributions.

This is data. Talking about data requires talk about statistics.

Cold fronts have strong temperature gradients. If your answer doesn't include the word 'gradient' then you can't possibly be answering the question I asked.

I have clearly and repeatedly explained what I think about these gradients whether they are real or not.

Yes, a 20 degree difference in temperature over 18 miles indicates a HUGE error in the thermometer record.

And 20 degree temperature differences are relatively rare in this data set. In the case of the two Iowa stations it represents something like 0.1% of the data. I am willing to assume that many if not most of those are just plain errors. Whether the recorder wrote the number down wrong or the instrumentation had problems.

Again, I am unwilling to toss out all the data if 0.1% of them have errors of that magnitude.

I am under the impression that the goal is to accurately measure the temperature is it not? If so, why doesn't someone actually try to figure out why 25% of the days have impossibly large temperature differences?

Again, I must point out that if this were a normal distribution then that would mean that 75% of the data is within +3 degrees, ergo the standard deviation would be less than 3 degrees.

No doubt, cold fronts goe through these two towns. But on 1 out of 4 days???? Are you serious?

I will again have to ask what kind of data sets you work with in your job. Do you have no error in your data? Do all the machines you use agree perfectly all the time such that there is no measurable difference at all between them?

Well if the error is greater than the purported warming you are in the silly position of saying that the world has warmed .84 +/- 4.07 degrees

I don't think that is how it is presented. I don't think that is how statistics works. In fact I have on a couple of occasions now explained how two different data sets can differ by an amount several times smaller than the standard deviation of either one.

I honestly don't know how much more explicit I can be on that point. If I have erred in the use of the t-test or the assessment of the data, please show me with actual calculations and equations.

(Hint: if you need some help on typing equations in CF, what I found useful is the "{sub}" and "{/sub}" and {sup} and {/sup} tags --replace the curly brackets with square ones and it allows you to type the equations. Like I've been doing on so many of these posts.)

I think this is possibly the most important part of the whole debate. You appear to be aware of some method of measurement that can span 6 decades (or more in some cases) and yet have no error.

Also, do keep in mind that a 1 deg F difference is about 2% of the average temperature for one of the Iowa stations 60-year average. The 60 year average temperature of the two stations is 48.5deg F, 1 degree difference is 2% of that amount.

The average of each station is:

Station #130600: 48.937deg F
Station #138296: 48.130deg F

 

[grmorton]

I guess that Contracelsus is leaving us ( I too hope he has a nice life as well). I thought I would post the temperature difference between Morrison Illinois and Clinton Iowa. You can see that up until 1975 Morrison was generally hotter (the red 365-day running average). But in 1975 the Clinton station moved by 1 mile, 5280 feet and for the next 30 years Clinton was hotter than Morrison. This illustrates how even small changes to the system make it difficult to measure tiny tiny temperature trends.

I know I am hated here (not that I really care cause we will never actually have beers together) but it is getting rather silent. Anyone want to say that the picture below is evidence of the weather service measuring the temperature trend properly?

 

[AV1611VET]

I know I am hated here...

Join the club, bro ---

 

[Contracelsus]

Given that you think this kind of error is ok, and given that you keep talking about distributions rather than in terms of physics and meteorlogy, I don't find that your 'counter points' are really even relevant to the question I am asking.

Unfortunately this is data we are discussing. It has always been data we are discussing. Remember, you were so happy when I first posted a histogram and started discussing the data. Well, this is discussing the data. From the scientists I've met this is precisely how data is dealt with.

Of course if I didn't appreciate how data behaves in the real world I'd be asking questions as if all data is real and there is never any error in it. But I don't expect that kind of stuff from one as knowledgeable as you in the areas of science.

If you have any questions about the statistics I am using here's a couple of books I've found really helpful:

Elementary Statistics in Social Research, 6th ed, Jack Levin and James Fox. Allyn and Bacon, Boston, 2000.

Introductory Statistics with R (2nd ed), P. Dahlgren, Springer, 2008

JMP Start Statistics: A guide to statistics and data analysis suing JMP (4th ed),John Sall, Lee Creighton, and Ann Lehman,SAS Publishing, 2007

I've also posted several websites in our discussion that outline how the t-test works as well as other useful information like the chi square test. In fact working with a lot of data as you must in your job in oil, there are probably many places you can find extremely useful information.

Temperature differences are detectable by the mere fact of subtraction. That is an interesting mathematical invention from about 3500 BC.

I am curious as to why you seem to be studiously avoiding the statistical treatment of the data. I think I have clearly laid out my calculations for all to see, so if there is an error I would welcome the critique, however I must be made aware how the statistical reasoning and calculation is in error.

I can even detect differences at the 20,000ths standard deviation by mere subtraction. If I have a data set with a std. dev. of 1, if two values are different by 6 standard deviations, then subtraction will give an answer of 6. If two of the points are separated by 20,000 sd, then the subtraction will yield 20,000. It isn't that hard Contracelsus.

I must not have been clear in my (now several) discussions of the t-test. This is relatively fundamental statistics. I thought I had made it quite clear that I have differentiated between two data sets who had a difference in means of 1 while each of their standard deviations was 6.

The important thing about the t-test is that it utilizes the standard error of the difference between the means. The term (often written as S[sub]x1-x2[/sub]) is not just the standard deviation of the two data sets, but rather the square root of these standard deviations with the sample number in each set factored in as multipliers etc.

Here's a link that explains the t-test:
The T-Test

The important bit in the posts where I showed this in action was a thing called the p value which is the probability of making an error in rejecting the null hypothesis (in the present case that there is no difference between the two data sets). In both of the examples I provided the p-value was very, very small. If p < 0.05 one can say with 95% confidence that they are probably not making an error in assuming the difference is not zero.

The hard thing comes from measuring a 1 degree change when the noise is 6x that

Not really with 20,000 pieces of data. (As I have shown in a couple of different posts).

I just told you what you were missing

And you did it without using a single statistics equation! In fact you appeared to have done it without talking about statistics at all

 

[Contracelsus]

I guess that Contracelsus is leaving us

Oh I am most assuredly not leaving! I am enjoying the statistics and discussion of data!

I know I am hated here

I hope I have not given you that impression by my posts!