Little Fish, on 07 December 2012 - 09:30 PM, said:
if I want to calculate the trend for the last 15 years, then I do not need the data from years previous to those 15. all I need is the data for the last 15 years, and I should not throw away any data as a declared "outlier". to throw away the el nino and not throw away the la nina's and el ninos that followed is fraudulent, so we should use all the data over the last 15 years.
But you still need 30 data points to calculate a statistically valid trend. With 15, you have a skewed result. Climate data is an autocorrelated time series - last year's data point is a better-than-average estimate of this year's. It lacks independence and because of that, a whole lot of calculating is needed to make the corrections. I just finished doing that with 482 tree ring series. It's a pain and I can understand why you don't want to do it. But if you want a relaiable result, you don't have a choice.
One other problem: a trend is a point-value. It only describes the rate of change at one point in time. The trend could change five minutes after you measured it. It could have been different (and probably was) seconds earlier. So even if you have one, it's not of much use.
Quote
I have explained the point to you SIX times now as to why we should look at 15 years of data. here for the 7th time - a flat trend of 15 years falsifies the gcms - that is what the gcm modelers and NOAA stated in 2008 - in order to assert that co2 is a scary life threatening primary climate driving gas that the gcms animate it as, the temperature trend will always show a rise over any 15 year period - their words, not mine.
Saying it again doesn't change the fact that your math is all screwed up.
Quote
the main temperature dataset used by the IPCC, HADCRUT3 and used in the graphs I showed is a monthly dataset, so 15 years of data contain 180 data points. the trend over the last 15 years is flat as I already showed you.
So now you've decided to introduce another set of variables to the problem. The problem that you're up against here is that by using monthly values, you have added a lot of variability to your dataset. And that widens your confidence interval. What you are tring to do is prove that the rate of slope is zero. But with a wide confidence interval, you can't disprove values that differ widely from zero. You have accomplished nothing.
BUT: By adding twelve dummy variables (one for each month) to the model, you can improve your accuracy considerably. That will give you 180-13=167 degrees of freedom. That should be enough. Did you do the math? What did it show?
Quote
no global warming for fifteen years falsifies the famous fear flooding fanatical fatalists fantastical fantasies.
You haven't disproven global warming during the 15 years in question, yet, but I'm anxious to see how you modeled the months. You might actually succeed. Time to show your work.
Doug
If I have seen farther than other men, it is because I stood on the shoulders of giants.
The beginning of knowledge is the realization that one doesn't and cannot know everything.