RSS

Dye 3 Temperature – Is it Normal?

24 Jan

Here is the Dye 3 temperature data used in the posts http://wattsupwiththat.com/2012/01/21/ap-index-neutrons-and-climate/ and http://wattsupwiththat.com/2012/01/24/the-message-in-the-dye-3-data/.  It has been detrended and normalized.  The bins are 0.01 sigma wide.  Actual frequency should be compared with normal density.  The accumulated actual and accumulated normal lie almost on top of each other.  The last series is the difference between (actual accumulated and standard normal accumulated)*10.  The tiny differences you see in the accum charts are multiplied by 10 so you can see which accum is above the other.  This is as normal as you are going to find.  I didn’t bother with any tests for normality because the result is obvious.

About these ads
 
2 Comments

Posted by on January 24, 2012 in Climate

 

2 responses to “Dye 3 Temperature – Is it Normal?

  1. Robin

    January 25, 2012 at 5:30 am

    I like this analysis. My own, posted in WUWT yesterday fully confirms your statement that the data are for all practical purposes normal. As you say, really no need for formal stats! However, I believe that the “linear” trend, whose stats I gave in my post and is clearly fine on the coarsest scale, in fact disguises some substantial and prolonged departures from linearity. Have you considered this?

     
  2. Michael D Smith

    January 25, 2012 at 2:23 pm

    I was thinking of doing a running test for differences to detect shifts. The problem with these distributions are that they are really random walks. Rodionov did some work for detecting regime changes, as an add-in for excel. It works in office XP, but not in office 2003. The technique is hidden so I would probably have to write something myself anyway so I know what it is doing, or find out more about his technique… If you search for Sergei Rodionov, you should be able to find the add-in if you want to try it.

    I have no doubt that the shifts are time dependent and are real and significant. The distribution as a whole looks like random normal with a trend but this doesn’t detect running changes. A good way to detect shifts would be using ranges between T and T-1, for example, to establish a short term sigma using d2, then detect when entire, larger groups are shifted away from a given trend line. Using control chart techniques could detect changes, the underlying statistics are similar.

     

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

 
Follow

Get every new post delivered to your Inbox.

%d bloggers like this: