Asymmetric uncertainty intervals in astrophysical data

Click For Summary

Homework Help Overview

The discussion revolves around calculating asymmetric uncertainty intervals in astrophysical data, particularly in the context of a machine learning project related to astronomical observations. Participants are exploring how to express uncertainty correctly and whether averaging upper and lower values is appropriate.

Discussion Character

  • Conceptual clarification, Assumption checking, Exploratory

Approaches and Questions Raised

  • Participants discuss the method of averaging upper and lower values to create a symmetric uncertainty interval and question its validity. There are inquiries about alternative techniques to achieve the desired uncertainty representation.

Discussion Status

The conversation is ongoing, with some participants providing guidance on the need to clarify the definition of "average" in this context. There is an acknowledgment of differing interpretations regarding how to handle uncertainty, but no consensus has been reached on a specific approach.

Contextual Notes

Participants are considering the implications of the authors' choices in expressing uncertainty and the potential need to reconcile their methods with established practices in the field.

ElectroFractal
Messages
11
Reaction score
0
Homework Statement
I have come across astrophysical data where the data have asymmetrical uncertainty intervals like this for example: 4.467[SUP]+0.064[/SUP][SUB]-0.096[/SUB]. I cannot realize how to calculate an average of this quantity, and a percent uncertainty.
Relevant Equations
I calculated the upper and lower value: 4.467+0.064=4.531 and 4.467-0.096=4.371, and found the average which is 4.451, and it's different from the value 4.467.
My initial guess was to calculate the upper and lower value, and then average those two values, but I don't know whether this is correct to make the uncertainty interval symmetric.

After I calculated the average value, I subtracted it form the upper and lower value, and obtained the symmetric interval: 4.451+-0.08. Thus, the percent uncertainty is:
0.08/4.451 = 1.79%

My question is if this way of thinking is a correct one, and if there are other techniques to obtain my goal?
 
Physics news on Phys.org
ElectroFractal said:
is a correct one
Obviously: No. Otherwise the authors would have published their result in that manner.

What is the motivation behind your 'goal' ?
 
My motivation is to find an average value of the observed quantity, and calculate a percent uncertainty. I am doing an ML project related to astronomical observations, and I am trying to do feature engineering to obtain better results from the model.
 
ElectroFractal said:
My motivation is to find an average value of the observed quantity, and calculate a percent uncertainty.

But the authors clearly did not think this was the correct way to express the uncertainty, otherwise they would have done it the other way.
 
So, is there a way to calculate what I want, or I should forget about this, and try to do something different?
 
You want the error to be symmetric. The authors didn't think it was. You need to decide how to handle this disagreement.
 
As @Vanadium 50 said, you need to think about what you are trying to do. What do you mean by the "average"? An average usually refers to reducing a list of values to a single number by adding up the values and dividing by the number of values. But here you have just a single value, so using the term average doesn't really make sense. Do you mean the most likely value? The authors have given you that. It is the 4.467 value.
 
  • Like
Likes   Reactions: ElectroFractal

Similar threads

  • · Replies 0 ·
Replies
0
Views
4K
  • · Replies 0 ·
Replies
0
Views
3K
  • · Replies 5 ·
Replies
5
Views
2K
  • · Replies 13 ·
Replies
13
Views
4K
  • · Replies 8 ·
Replies
8
Views
6K
  • · Replies 12 ·
Replies
12
Views
4K
Replies
5
Views
5K
Replies
67
Views
8K
Replies
1
Views
2K
  • · Replies 76 ·
3
Replies
76
Views
7K