Gust of Hot Air is a blog outlining my own statistical analysis of Australian Weather. I am Jonathan Lowe, and have completed by Bsc(hons) in statistical analysis as well as my Master of Science. I have done 2 years of my PhD
There is a lot of statistical information regarding climate change and I intend to provide statistical analysis into the area to prove if the recent well advertised rise in temperature is at all statistically significant. Results will be uploaded here on a regular basis
Friday, November 03, 2006
Walk Against Warming
Australia are officially walking against warming on Saturday.
I argue that a bit of research might actually do them better. Maybe for instance....reading this blog perhaps....
15 comments:
Anonymous
said...
Hi Jonathan. Great Blog you have here. I only wish I had half the scientific credentials you have. At least I have critical thinking, and it tells me that CO2 is not a driver of climate.
Hi Jonathon. I only wish you had scientific credentials in climatology. I have critical thinking, and it tells me that CO2 is a driver of climate, and your blog proves nothing.
In fact I will start my own blog to prove this, and I will leave a comment with a link to it.
thanks for your reply umm...Anonymous. In fact, if I was a climatologist, then I would not be able to analyse the temperature data anywhere near the sophistication that I can analyse it as a statistician. That is after all, what statisticians are the best at.
Sophistication? All you've shown so far is that you don't see an increase in average temperatures taken from a limited number of weather stations in Australia.
What does that prove? That it is highly unlikely that average temperatures have increased by more than 2 °C in the last 100 years in Australia? But that's consistent with the data of climate scientists (which indicates that the average global temperatures has increased by 0.6 °C).
Also, global warming leads to an increase of the average global temperature, not necessarily to particular regions of the world.
So, your work has no relevance to climate research.
Hi Count Iblis, A limited number of stations? I have chosen about 100 stations scattered australia wide that show long term temperature readings. Hardly limited.
"What does that prove? That it is highly unlikely that average temperatures have increased by more than 2 °C in the last 100 years in Australia?"
Thanks for trying to put words in my mouth but no, that Australia has not increased temperature by 0 degrees celcuis. There has been no increase at all in Australia. Not even 0.5 degrees.
My analysis is from Australia only, not the world. It could well be warming up in Canada, and there might be a heat wave in Tanzania, but the evidence suggests no warming in Australia. Hence no need for Australians to be majorly concerned, and no proof that Co2 and human interaction has heated the country.
Global warming should therefore be renamed partial-global warming.
"Thanks for trying to put words in my mouth but no, that Australia has not increased temperature by 0 degrees celcius. There has been no increase at all in Australia. Not even 0.5 degrees."
What are the error bars?
"My analysis is from Australia only, not the world. It could well be warming up in Canada, and there might be a heat wave in Tanzania, but the evidence suggests no warming in Australia. Hence no need for Australians to be majorly concerned, and no proof that Co2 and human interaction has heated the country."
This is just Straw Man research. You have (attempted to) disproved a point that is not (necessarily) a feature of global warming scenarios.
The dangers posed by Global Warming for Australia is certainly not the mere fact that temperatures could become a few degrees warmer. But some climate models show that El Niño could become a permanent feature leading to severe droughts.
Hi Count Iblis, no offense, but error bars are rarely used by statisticians because they are just too plain simple and don't really prove anything (we learnt error bars at 1st year uni and never touched it after that). We prefer to use ARMIA models, regressional analysis and significance tests, which better analyse the data rather than just including error bars on graphs.
I have proven that Australia is not warming, is not facing increased wild weather, and it not facing increased droughts. What other effects of global warming (or perhaps I should refer to it as partial-global constant temperature) are so hazardous to spend godzillions on?
what I mean by error bars is the confidence limit. I don't really care that much what kind of statistical methods you've used, I'm assuming that what you've done is correct.
You say that:
"I have proven that Australia is not warming, is not facing increased wild weather, and it not facing increased droughts."
Within what limits? What is the confidence interval? I mean you cannot have concluded that Australia is not warming or cooling at all literally, as that would imply that you've ruled out a temperature change by even a million-th of a degree °C .
Count Iblis, when we use scientific tests we set out to prove whether or not we hace achieved a significant increase or decrease in temperature. Eg. has the temperature increased or decreased not due to random variation. The p values that you see listed are the probabilities that the increase or decrease is due to chance. Generally this has to be less than 0.05 for us to conclude significant increase/decrease in temperature. But if you want error bars, the most recent 9pm analysis from 1960 onwards suggests a 0.0046 degree ceclius increase per year +/- 0.0076 which is statistically insignificant.
"But if you want error bars, the most recent 9pm analysis from 1960 onwards suggests a 0.0046 degree ceclius increase per year +/- 0.0076 which is statistically insignificant."
Ok, let's work with this figure. You are 95% sure that any trend is between (0.0046 +/- 0.0076)°C/year = (0.46 +/- 0.76)°C/century.
(Actually, you cannot say 95% sure, for that you must assme a uniform prior on the possible trends.)
The figure given by climate scientists for the temperature increase in the last 100 year lies within your bound.
That's why your analysis is not relevant. It's simply not accurate enough to see the signal claimed by climate scientists.
You can always assume a null hypothesis that says that there is no change, do an analysis that can only yield a significant result rejecting that hypothesis if you have a huge effect and then claim that just because that isn't the case, the null hypothesis is not overruled.
The null hypothesis is already overruled by more accurate research involving many thousands of weather stations and satellite data.
Well if you want to throw the entire method of statistical research out the door and conclude what you want to make of the results to suit your need then go ahead. But the results are insignificant. There is no evidence to prove an increase or decrease in temperature. But feel free to spend gozillions of dollars on something that hasn’t been proven let alone proven of mans influence.
But if you want to go against the method of science, then here are the results for the times done on the webpage so far:
Hardly conclusive, and in fact, let me remind you again: . There is no evidence to prove an increase or decrease in temperature. Insignificant evidence.
damn I stuffed up. Ok for 9pm the coeffecient is 0.004591 with the standard deviation 0.003750, or over 100 years 0.46 +/- 2* 0.375 = 0.46 +/- 0.75. Sorry about that.
15 comments:
Hi Jonathan. Great Blog you have here. I only wish I had half the scientific credentials you have. At least I have critical thinking, and it tells me that CO2 is not a driver of climate.
Keep it up, Mate!
Hi Jonathon. I only wish you had scientific credentials in climatology. I have critical thinking, and it tells me that CO2 is a driver of climate, and your blog proves nothing.
In fact I will start my own blog to prove this, and I will leave a comment with a link to it.
Jonathan,
I did't think sheep could read...maybe thats why they can't research the facts!
Cheers
thanks for your reply umm...Anonymous.
In fact, if I was a climatologist, then I would not be able to analyse the temperature data anywhere near the sophistication that I can analyse it as a statistician. That is after all, what statisticians are the best at.
Sophistication? All you've shown so far is that you don't see an increase in average temperatures taken from a limited number of weather stations in Australia.
What does that prove? That it is highly unlikely that average temperatures have increased by more than 2 °C in the last 100 years in Australia? But that's consistent with the data of climate scientists (which indicates that the average global temperatures has increased by 0.6 °C).
Also, global warming leads to an increase of the average global temperature, not necessarily to particular regions of the world.
So, your work has no relevance to climate research.
Hi Count Iblis,
A limited number of stations? I have chosen about 100 stations scattered australia wide that show long term temperature readings. Hardly limited.
"What does that prove? That it is highly unlikely that average temperatures have increased by more than 2 °C in the last 100 years in Australia?"
Thanks for trying to put words in my mouth but no, that Australia has not increased temperature by 0 degrees celcuis. There has been no increase at all in Australia. Not even 0.5 degrees.
My analysis is from Australia only, not the world. It could well be warming up in Canada, and there might be a heat wave in Tanzania, but the evidence suggests no warming in Australia. Hence no need for Australians to be majorly concerned, and no proof that Co2 and human interaction has heated the country.
Global warming should therefore be renamed partial-global warming.
"Thanks for trying to put words in my mouth but no, that Australia has not increased temperature by 0 degrees celcius. There has been no increase at all in Australia. Not even 0.5 degrees."
What are the error bars?
"My analysis is from Australia only, not the world. It could well be warming up in Canada, and there might be a heat wave in Tanzania, but the evidence suggests no warming in Australia. Hence no need for Australians to be majorly concerned, and no proof that Co2 and human interaction has heated the country."
This is just Straw Man research. You have (attempted to) disproved a point that is not (necessarily) a feature of global warming scenarios.
The dangers posed by Global Warming for Australia is certainly not the mere fact that temperatures could become a few degrees warmer. But some climate models show that El Niño could become a permanent feature leading to severe droughts.
Hi Count Iblis,
no offense, but error bars are rarely used by statisticians because they are just too plain simple and don't really prove anything (we learnt error bars at 1st year uni and never touched it after that). We prefer to use ARMIA models, regressional analysis and significance tests, which better analyse the data rather than just including error bars on graphs.
I have proven that Australia is not warming, is not facing increased wild weather, and it not facing increased droughts. What other effects of global warming (or perhaps I should refer to it as partial-global constant temperature) are so hazardous to spend godzillions on?
Jonathan,
what I mean by error bars is the confidence limit. I don't really care that much what kind of statistical methods you've used, I'm assuming that what you've done is correct.
You say that:
"I have proven that Australia is not warming, is not facing increased wild weather, and it not facing increased droughts."
Within what limits? What is the confidence interval? I mean you cannot have concluded that Australia is not warming or cooling at all literally, as that would imply that you've ruled out a temperature change by even a million-th of a degree °C .
Count Iblis, when we use scientific tests we set out to prove whether or not we hace achieved a significant increase or decrease in temperature. Eg. has the temperature increased or decreased not due to random variation. The p values that you see listed are the probabilities that the increase or decrease is due to chance. Generally this has to be less than 0.05 for us to conclude significant increase/decrease in temperature. But if you want error bars, the most recent 9pm analysis from 1960 onwards suggests a 0.0046 degree ceclius increase per year +/- 0.0076 which is statistically insignificant.
"But if you want error bars, the most recent 9pm analysis from 1960 onwards suggests a 0.0046 degree ceclius increase per year +/- 0.0076 which is statistically insignificant."
Ok, let's work with this figure. You are 95% sure that any trend is between (0.0046 +/- 0.0076)°C/year = (0.46 +/- 0.76)°C/century.
(Actually, you cannot say 95% sure, for that you must assme a uniform prior on the possible trends.)
The figure given by climate scientists for the temperature increase in the last 100 year lies within your bound.
That's why your analysis is not relevant. It's simply not accurate enough to see the signal claimed by climate scientists.
You can always assume a null hypothesis that says that there is no change, do an analysis that can only yield a significant result rejecting that hypothesis if you have a huge effect and then claim that just because that isn't the case, the null hypothesis is not overruled.
The null hypothesis is already overruled by more accurate research involving many thousands of weather stations and satellite data.
Well if you want to throw the entire method of statistical research out the door and conclude what you want to make of the results to suit your need then go ahead. But the results are insignificant. There is no evidence to prove an increase or decrease in temperature. But feel free to spend gozillions of dollars on something that hasn’t been proven let alone proven of mans influence.
But if you want to go against the method of science, then here are the results for the times done on the webpage so far:
9am: -0.4 +/- 0.08 (wow a significant decrease!)
3pm: -0.1 +/- 0.1
9pm: 0.45 +/- 0.04
3am: -0.1 +/- 0.2
Hardly conclusive, and in fact, let me remind you again: . There is no evidence to prove an increase or decrease in temperature. Insignificant evidence.
How did these confidence intervals get smaller by a factor of ten?
hmm no they haven't, can you explain? think you misread
damn I stuffed up. Ok for 9pm the coeffecient is 0.004591 with the standard deviation 0.003750, or over 100 years 0.46 +/- 2* 0.375 = 0.46 +/- 0.75. Sorry about that.
Post a Comment