Systematic Redshift: Exploring Mg II Uses

[shadishacker]

Hi everyone,

I am new to observations and observational terms!
I am reading the paper "constraining the time variation of the fine-structure constant" by Srianand et. al
in the section "constraining alpha with QSO absorption lines" there is a sentence saying "... rest wavelengths of MG II ... thereby providing a good anchor for measuring the systematic redshift."

- What does the systematic redshift mean?
- What do they use Mg II for?

 

[Bandersnatch]

- What does the systematic redshift mean?

The redshift due to the expansion, as opposed to what they want to measure - i.e. changes in alpha.

- What do they use Mg II for?

You want to first identify transitions that would not vary much with changes to fine structure constant, and use these as the baseline for deducting the redshift due to expansion from your observables. The changes in transition energies that are left then can be interpreted as due to varying alpha.

 

[shadishacker]

Thanks a lot for the clear reply.
I'd like to ask two other questions:

- About this variation, in recent papers like " High-precision limit on variation in the fine-structure constant from a single quasar absorption system " by Kotus et. al mention that the measured variation for alpha is consistent with no variations.
But by the limits they say -1.4\pm 0.55 \pm 0.65 ppm, the term "no variation" is not fixed! I mean there are constraints on varying alpha, that 0 is one of them; but still it is not rejected.
As I right?

- In "CONSTRAINING THE VARIATION OF THE FINE-STRUCTURE CONSTANT WITH OBSERVATIONS OF NARROW QUASAR ABSORPTION LINES" by Songaila et. al, they used the sentence:
"We find a null result of ..."
what do they mean? They are reporting the probability of small variations in alpha, but they are using the term null result!
What does it mean?!

 

[Bandersnatch]

they are using the term null result!
What does it mean?!

It means that the result does not support the hypothesis. Here, the hypothesis is something like 'alpha varies with time'. If your experiment, designed to find this variation, nets you a range of possible values that includes 0 (no variation), and is not statistically leaning to one side (e.g. 0 is at the extreme end of the range), then your experiment has failed to find evidence of the proposed variation.
Alternatively, you can say that your experiment is consistent with the null hypothesis - here, the null hypothesis is that alpha does not vary in time.

This methodology is used in all science. It's simply a codified way of following the rule that 'you need extraordinary evidence for extraordinary claims'. A null result is a result of no consequence (which doesn't mean it was a waste of time!).

I believe this also answers your other question.

 

[shadishacker]

But how this "null result" given in this number? $\Delta \alpha/ \alpha = (-0.01\pm0.26) \times 10^{-5}$ which is for instance given in CONSTRAINING THE VARIATION OF THE FINE-STRUCTURE CONSTANT WITH OBSERVATIONS OF NARROW QUASAR ABSORPTION LINES" by Songaila et. al ?
I mean, how is 0 at the extreme end of the range?

 

[Bandersnatch]

But how this "null result" given in this number? $$\Delta \alpha/ \alpha = (-0.01\pm0.26) \times 10^{-5}$$ which is for instance given in CONSTRAINING THE VARIATION OF THE FINE-STRUCTURE CONSTANT WITH OBSERVATIONS OF NARROW QUASAR ABSORPTION LINES" by Songaila et. al ?
I mean, how is 0 at the extreme end of the range?

It isn't. That's the point. It's smack dab in the middle of the range. Hence you can't conclude that you've found evidence of variability. The result does not support it (= it's null).

 

[shadishacker]

It isn't. That's the point. It's smack dab in the middle of the range. Hence you can't conclude that you've found evidence of variability. The result does not support it (= it's null).

Am I missing something? Isn't the middle of the range -0.01?

 

[Bandersnatch]

Yeah, with error bars +/- 0.26.

 

[shadishacker]

Yeah, with error bars +/- 0.26.

So zero is not smack dab in the middle of the range, but it is -0.01. right?
do they conclude the variation is zero because of -0.01 is a small number near 0?

 

[Bandersnatch]

They don't conclude the variation is 0! They conclude that they found no evidence that it isn't.

And of course -0.01 and 0 are not the same.

Perhaps there's a thread level mismatch. You've marked this thread as A = graduate level. Haven't you taken statistical methods? Done some lab work?

Look, if they found the result being e.g. -0.010 +/- 0.002, then they would have found evidence of variability. Same if it were e.g. 0.80 +/- 0.26. This is because both would exclude the possibility of there being no variation.
But the results they did get do not exclude 0, nor do they exclude anything between 0.25 and -0.27. The actual value can be anything in that range, with values around -0.01 being the most likely. Somebody would have to devise a better experiment to narrow the error bars even more, but as long as the new experiment results include 0, the results will be null.

An in case this is also causing confusion, 'null result' does not mean 'equal to zero'. It means 'of no consequence', or 'not sufficient to support the hypothesis'.

 

[shadishacker]

Dear Bandersnatch,
Sorry if I marked the level wrong.

Anyway, thanks for the explanations.
I get the point now.