Dear Tribhuwan Singh Bisht

Try to compare the standard errors of your parameters, which takes the sample size into account..

Let's start at the beginning. What do you want to find out?

David Booth

Dear Tribhuwan Singh Bisht ,

If you have access to the raw data you can create an error bar chart which will enable you to compare the means of several "parameters" with their associated standard errors. Non-overlapping intervals indicates likely significant differences.

David Eugene Booth Sir

My actual problem is based on Delphi technique

1. There are 30 experts and 50 parameters and the experts are asked to rate these parameters on a scale of 1-5.

2. The experts may choose to rate the particular parameter or not.

3. So for these 50 parameters i will be having different number of experts rating each parameters.

4. I have to select 20 most important parameters based on the experts ratings, but the sample size ( the number of experts rating a parameter is different for each parameter) suppose parameter 1 is rated by 50 experts whereas parameter 2 is rated by 35 experts, parameter 3 is rated by 45 experts and so on.

5. So the main problem here is how do i select the 20 most important parameters. If i go by the average of the ratings than it won't reflect the actual case as sample size is different.

6. Which method should i use in order to compare different parameters so i could select 20 most important parameter out of the 50 parameters that i have.

Masoud M. Hassan ,

Thank you for your answer but the magnitude of the standard error gives an index of the precision of the estimate of the parameter. How would it help in comparing or ranking the parameters.

Look this is Fun but if these are really parameters not estimates then the answer to your original question was subtraction.. There is no mention of a sample anywhere. If This is a rater- choice problem it likely belongs to graph theory, like the birthday problem. After rereading My best advice is take a graph theory course. Don't you have real research to ask about.? Or is this really a graph theory take home exam.? David Booth

David Eugene Booth

It's about research only i am trying to develop an environmental index using Delphi technique. I have the responses of the experts on the 50 parameters. Now for the final index i want to reduce the number of parameters to 10 only so that it is easy to use, understand and estimate. Now based on the ratings given by the experts which is the most suitable way to rank the parameters because in traditional Delphi technique we simply average the ratings and then decide upon the parameters that should be included in an Index. But, i consider this is not the best method as the number of experts giving their rating for a particular parameter are different for different parameters.

OK there is a whole literature on this rating system. Much of it is at this link:

https://www.google.com/search?q=combining+raters+ratings+from+Delphi+method&rlz=1C1CHBF_enUS847US847&oq=combining+raters+ratings+from+Delphi+method&aqs=chrome..69i57j33.33272j1j7&sourceid=chrome&ie=UTF-8

See if there is something you can use here. Let me know the result. I'll keep looking. Best wishes, David Booth some advice: if people are trying to help you, do not hide information from them.

David Eugene Booth

I wasn't trying to hide information in fact i was trying to ask my question in the simplest way possible. I am sorry if you felt that i was trying to hide information from you. I will go through the link you have provided and will revert back to you once i am done. Thank you for your help.

David Eugene Booth

I have gone through various literature and the for the rating system median or mean method is being used.