Experimental Conditions (modeling)

[caitlin]

Moved from ​http://www.ome-xml.org/ticket/102

I’m concerned about the modeling of the experimental conditions, particularly the Experiment element.

1. First: An Experiment element can have multiple Images. So, how can Image-dependent characteristics, like the type of microscopy used to collect the image, be modeled as Experiment-dependent attributes, like Experiment/Type? as it is currently recorded? This seems like a normalization problem. Are you counting on your users to only group Images that share the same kind of microscopy? I don’t think you should.
2. What the word “experiment” means will vary significantly from user to user, and even for a given user at different times, to mean single acquisition events, multiple closely related acquisition events, or even other things. I think you are inviting confusion and error as a side-effect of using a word with established but informal meanings, like experiment, to mean formal but locally-invented things, like “collection of datasets.” Users (and developers) end up using the word experiment intuitively, and then you have mix-ups.
3. The main thing Experiment really adds is a redundancy—another way to group Images together. I recommend that you eliminate Experiment, and simply continue to allow users to group images as Datasets and Projects. If a Dataset or Project represents a single experiment for that user, great!—they can use whatever organizational system suits them, without us committing any additional metadata to it.
4. As for the rest of the experimental conditions, there are several ways you could incorporate them:
   1. They could simply be added as subelements or attributes of Image. This has the benefit of simplicity.
   2. The subelement of Image called “ImagingEnvironment?” could be used as an umbrella to collect all these experimental conditions, instead of just the four conditions it holds right now. In addition to Experiment Type and MicrobeamManipulationRef?, I would also include ObjectiveSettings?. This has the benefit of making it easy to see where new experimental conditions should be added, if you wish to extend them in the future.
   3. You could create a new subelement of Image to do essentially the above, if you did not wish to expand the use of ImagingEnvironment?.