Originally posted by JonathanLowe:
Count Iblis II, can I ask you what your credentials are in Statistical analysis? No offence, but doing a monte carlo analysis based on your assumptions of what you think the weather stations should say based on climate change and then making 95% confidence intervals based on these simulations I believe is absurd.
Better than that. Why not get the data, test the data for trend analysis. Who needs monte carlo simulations for data that is already there, of which should you do them is subject to what you think should happen in the first place. This is not good statistical analysis.
Absurd? I don't think you get it. How can you conclude anything based on your measurements? What are the error bars. What Global Warming scenarios are constrained given your conclusions?
The Monte Carlo method is necessary to validate your method. If you don't do it, no one will take you serious. It is standard practice in many scientific fields ranging from astrophysics, particle physics etc. The people at CERN who will soon be swamped with thousands of terabytes of data are busy doing simulations using the software that they'll later use to extract properties of particles they hope to discover when the Large Hadron Collider becomes operational. They need to be sure that if they see a Higgs particle, it is indeed a Higgs particle and not some software bug, or perhaps some sublte flaw in the reasoning that led to a flawed algorithm used in the data analyses.
They run simulations of the detectors themselves to generate the data that would be produced according to certain theoretical scenarios. They then run their data analysis software on the (fake) data to see if what comes out using is indeed consistent with what they put in.
If you only rely on "theory" to validate your data analysis technique you are unlike to spot any subtle flaws. In your case, no one will have any confidence that, if there is a trend toward higher average temperatures consistent with what most climate scientists think is going on, you would able to pick it up using your method.
Average global temperatures have only increased by 0.6 degrees C over a period of more than 100 years. That's a small increase over a large period and will be completely swamped by natural fluctuations at indivudual stations. The signal only becomes visible after averaging over a large number of stations.
