## CMB data help

Zhao Wen
Posts: 5
Joined: March 20 2005
Affiliation: USTC Hefei China

### CMB data help

Dear friends:
Who can explain the meaning of binned data of CMB power spectra? How to get the binned data (mean value and the error bar) from the unbinned data? or introduce a paper for me? Thanks a lot!

Sarah Bridle
Posts: 144
Joined: September 24 2004
Affiliation: University College London (UCL)
Contact:

### CMB data help

I thought the binned data was just the weighted mean of the unbinned data (with the weighted mean error bar)? It certainly makes for a sensible looking plot..

Zhao Wen
Posts: 5
Joined: March 20 2005
Affiliation: USTC Hefei China

### Re: CMB data help

Thank you very much! But, as we known, in the bined data, the error bar is much smaller than the unbined data. If is only the weighted mean, the error bar of binned data can not be obtained.
Sarah Bridle wrote:I thought the binned data was just the weighted mean of the unbinned data (with the weighted mean error bar)? It certainly makes for a sensible looking plot..

Anze Slosar
Posts: 183
Joined: September 24 2004
Affiliation: Brookhaven National Laboratory
Contact:

### Re: CMB data help

Zhao Wen wrote:Thank you very much! But, as we known, in the bined data, the error bar is much smaller than the unbined data. If is only the weighted mean, the error bar of binned data can not be obtained.
Sarah Bridle wrote:I thought the binned data was just the weighted mean of the unbinned data (with the weighted mean error bar)? It certainly makes for a sensible looking plot..
If you add 1/\sigma^2 it should probably be good enough for most purposes...., i.e. 1/\sigma_binned^2 = \sum 1/\sigma_i^2

Zhao Wen
Posts: 5
Joined: March 20 2005
Affiliation: USTC Hefei China

### Re: CMB data help

Thank you very much! I have known it!

Anze Slosar wrote:
Zhao Wen wrote:Thank you very much! But, as we known, in the bined data, the error bar is much smaller than the unbined data. If is only the weighted mean, the error bar of binned data can not be obtained.
Sarah Bridle wrote:I thought the binned data was just the weighted mean of the unbinned data (with the weighted mean error bar)? It certainly makes for a sensible looking plot..
If you add 1/\sigma^2 it should probably be good enough for most purposes...., i.e. 1/\sigma_binned^2 = \sum 1/\sigma_i^2