Definition of "noiseVarianceHomogeneous" -> "errorVarianceHomogeneous"

[cmaumet]

Proposed definition:

nidm:noiseVarianceHomogeneous: "A boolean value that indicates whether the variance of the noise is modelled as being constant throughout the observations used in the model parameters estimation."

(This issue is a companion for #176).

[nicholst]

What about this:

nidm:noiseVarianceHomogeneous: "A boolean value regarding how the variance of the noise is modelled during parameter estimation; TRUE for constant variance over all observations in the model, FALSE for heterogeneous variance."

[cmaumet]

+1

[cmaumet]

Maybe to be consistent with #193, we could use "variance of the error" instead of "variance of the noise"?

[nicholst]

+++1! In fact, I'd be for making a wholesale change, Noise->Error, for all of the "Noise" terms listed in #176 @ elsewhere (e.g. GoogleFight "Error Variance" vs "Noise Variance")

Though I think the language is "Model for the (random) error"
instead of "Noise Model".

What do people think?

[cmaumet]

+1 for changing "noise" in "error" in all the "Noise Model" terms.

[khelm]

+1, I guess. i'm basically persuaded by the discussion above and agree that "error" is more inclusive than "noise". I'm more used to hearing about "modelling the noise", though that just may be due to my background and a Google search shows that a number of the top hits are for modeling of actual (acoustic) noise - which is also a point in the favor of "error".

[cmaumet]

Great! The following is now implemented:

nidm:errorVarianceHomogeneous: "A boolean value regarding how the variance of the error is modelled during parameter estimation; TRUE for constant variance over all observations in the model, FALSE for heterogeneous variance."

I positioned the curation status to "Pending final vetting". @khelm: if you also agree with this definition I think we could move to "Ready for release". What do you think?

[khelm]

Could we change "regarding" to "reflecting"?

[cmaumet]

+1

[nicholst]

+1 for reflecting

[cmaumet]

Thank you @nicholst and @khelm.

The following is now implemented and ready for release:

• nidm:errorVarianceHomogeneous: "A boolean value relecting how the variance of the error is modelled during parameter estimation; TRUE for constant variance over all observations in the model, FALSE for heterogeneous variance."

[khelm]

Wait I thought it was going to be: "A boolean value reflecting how the variance of the error is modeled during parameter estimation; TRUE for constant variance over all observations in the model, FALSE for heterogeneous variance."
*reflecting" rather than "regarding"
also "modeling" without the extra "l"

[khelm]

Sorry, I meant "modeled"

[cmaumet]

Hi @khelm, you are right it is "reflecting" not "regarding" (the implementation was correct). I have just updated my comment above.

I have just committed to master to correct "modelled" in "modeled", thank you for pointing this out.