There is increasing interest in analysing huge datasets and in applying “artificial intelligence”/”machine learning” methods to such analyses rather than more traditional statistical methods. As yet this has had relatively little development in therapy data but it will surely come and some work on applying these methods is emerging. We’re sure there are gains from collection of very large aggregated datasets, however, we’re more wary of much of the excitement generally about “big data” and “AI/ML” methods!
Details #
Not yet! However, this surely will develop and we’ll try to keep aware of developments.
Quick update 29.ii.24: interesting to see the above. I think the glossary has moved from “we” to “I” as J has had no real time for this (we always knew that!). Less parochially, there has indeed been a very rapid growth in application of various AI/ML methods to psychological therapy data where large datasets exist hence so far mainly from UK researchers working with UK IAPT datasets and large university datasets (so far only from Trier?) from mainland Europe. Am I failing to see North American work in this mode, the datasets must be there. There’s some reassuring work in the UK showing that the differences between the various ML/AI methods for some analyses were small and that, for those analyses, simpler methods and designs were at least as good as more complex ones.
My sense now is that this arena is in a position not unlike that of the application of multi-level modelling (MLM/HLM) about 30 years ago and will bring some useful findings slowly but I’m not yet convinced that the apparent strengths of the methods will have the impact that MLM has had. However, that’s perhaps because I’m biased toward models I can understand fairly easily and see as bringing in an improvement in the basic models, bringing viewpoints, that can be explained fairly easily. So far my reading about ML/AI really doesn’t suggest to me that it will ever have that clarity. That’s not to say it’s not powerful of course, it is!
Try also #
Cell size
Referential data
Statistical power
Precision
Estimation
Internal vs. external analyses
Chapters #
Rather paradoxically, only in Chapter 7 (p.116/117)!
Dates #
Created 14/11/21.
