Refactor batch CI generation (deprecates batchGenerateCI)

[aucuparia]

batchGenerateCI should be deprecated. generateCI should be augmented to add support for batch generating CIs. This has already mostly been done by adding the participants parameter.

Todo:

• Add autoscaling parameter to generateCI for use with the participants parameter.
• Rewrite batchGenerateCI to be a wrapper around generateCI.
• Add a "function deprecated" warning message to batchGenerateCI.

[rdotsch]

This should become a valid line of code:

generateCI(stim, response, "base", rdata, participants=subnum, individual_scaling="dependent", save_individual_cis=TRUE)

Importantly, individual_scaling="dependent" should be the new name for autoscale.

Two issues:

1. This is the scaling that should be applied to each individual ci, rename the group level CI scaling to group_scaling?

2. We should add another argument that specifies condition, with multiple participants nested in condition. If specified, the group-level images should be CIs per condition, and the group-level scaling argument should refer to these group-level images per condition then.

This has very high priority. @aucuparia, do you have time to implement this before the end of next week?

[rdotsch]

A third issue is the return value: it should return a list of cis, but because this can either be a list of subject-level cis, or group-level cis, we should probably return a list of lists, with two elements which are lists themselves. One element contains a list of subject-level cis, the other a list of group-level cis.

[rdotsch]

Alternatively, we can take the whole scaling issue out of the generateCI function and use the return value for generateCI in a new scaling function with more functionality than the autoscale function. What do you think?

[aucuparia]

Okay, I read your comments. Seems doable.

Regarding your question: I think I prefer to have a single function (generateCI) over having two or more (generateCI + scaleCI + ...).

We don't want to overwhelm novice users with lots of steps that have to be performed manually. We should make it as easy as possible for them to get the CIs. Ideally, they'd import their dataset, run generateCI with as few parameters as possible (because we set reasonable defaults where possible) and get their CIs.

Expert users can still get the raw or intermediate data by directing the output of generateCI to an object.

[rdotsch]

Alternatively, we can separate generateCI + scaleCI + ... for our own sanity, so the code stays manageable, and then write a wrapper to create a single function for novices.

[aucuparia]

@rdotsch, not that I object to separating generating and scaling into separate functions, but then what would the wrapper be called?

I'm reorganizing the code in generateCI now. I'll let you know when that's done. If after reviewing that you still prefer separate functions, we'll go with that.

[rdotsch]

How about generateAndScaleCI() orvisualise()?

In any case, let's indeed wait for your reorganisation first.

[rdotsch]

Let's stick go generateCI() for both generation and scaling, because we want to stick to the specification we used in the European Review.