Scientific Models vs Data Analytics.

[Aaron Hepi]

Has anyone bothered to think about the differences or similarities between the two? Their relative descriptive capacities? Their strengths and weakness? I imagine there is some kind of philosophy that illuminates all of the qualities I have described above; however, I can not find it.

Another interesting question is whether you think Data Analytics technology will reduce our dependence on the formulation of formal scientific models? By formulation, I mean humans doing the formulating, not machines?

Sorry if this all sounds very pie in the sky. I have no sophisticated interpretation of any of the above. So if anyone could point me in any direction at all, I would be grateful!

 

[Faye]

I thought data analytics meshed with science. It is all just advanced statistics, and they used advanced statistics in science.

 

[sprinkles]

A scientific model is testable. e.g. with fluid mechanics you can take fluids and experiment with them to show that the model works.

Data analysis is just analyzing data. For example you might watch the temperature of a fluid over time but it's just looking at numbers. Without another context it doesn't really tell you why the numbers are doing whatever. Data analysis might be used in scientific models. It's also often used in statistics, which are a bit more observational and maybe conjectural.

 

[Flavus Aquila]

Data analytics may be useful to find correlations; connections between two or more things. The more advanced the analytics, stronger correlations can be found.

For example, living in close proximity to bracken plants has some weak correlation with leukemia - but if you could crunch more data: genetic markers; exposure to bracken; respiratory infections; injuries; exposure to other chemicals; vaccination histories; etc.; etc. - you might start finding stronger correlations, which then allows better postulation about whether there is a cause-effect relationship - and the nature of that cause-effect. This in turn provides better models for scientific testing.

So how analytics and Sci models mesh - is just an expansion of observation, looking for causal connections between disparate things; and computerised analytics allows one to draw more disperse and multi-causal factors into consideration, while filtering for actual single/multi factor correlations. This allows one to target hypotheses and scientific research/experimentation to actually determine/confirm both the actuality and nature of causal effects in a far more complex model, than simple observation allows.