Quote:
Originally posted by JonathanLowe:
Hmm no idea how monte carlo simulations has anything at all to do with temperature analysis, but anyway.

I can admit that some people will obviously respect Climate scientists who are working int he field over me, just some blogger (a blogger who has spent 9 years at university studying statistcs by the way), and I can understand that people are skeptical of what I have found, however....

I will be publishing on my website how exactly I came to the conclusions and exactly what methods I used, so that, should you wish, you can also get the data from the ABM and replicate exactly what I've done in the way that I suggested and prove to yourself, that my analysis is unbiased and accurate.
Besides Dan's example, you need to prove that your method works. If people use methods for data analysis that are not standard or perhaps they are standard but applied in a slightly diferent way than usual, then you need to validate your method. There can be small subtle effects that can affect the outcome of the analysis. So, you need to test it just like you would test software to debug it: by doing experiments where you input data for which the outcome is known.

So, you treat your data analysis method as a black box. Data comes in and results come out. You then simulate the data that you would expect from weather stations for different climate change scenarios. This is where the Monte Carlo method comes in. The fake data is the trend plus local variability which is random on various time scales.

Using the simulations you can see how good your method is. E.g. at what rate must average temperatures increase for your method to detect the increase with 95% probability?