MDMWG report progress

It's not a bad place.

• The survey is finished
• We have (some) data
• It's time to think about the future!

So:

Conceptually, the final report has two main parts:

• A discussion of best practices
• Presentation of our findings

But what about the executive summary, you ask?

There are other parts to the report.

• Dr. Rials covered these quite well in outline at the last meeting!
• Most of them act in support to or summary of those I mentioned.

So I'd like to cover just the two I mentioned in a bit more depth!

The questions we've asked mostly count things or place them in categories.

We've also chosen a fairly high level of analysis

• Mostly, it's the agency as a whole, though there are exceptions.

This was to get results.

• The (impractical) perfect is the enemy of the (practical) good!

This limits the formative use of our analysis!

Most causal and even relational analyses are ruled out (or very complicated).

Counts and categories are useful to get the lay of the land.

But they don't explain or show a way forward!

That's not good or bad – it's just the nature of the data.

Some of the more complicated analyses might be useful in-house.

But I'm not sure we want to present high-dimensional regressions here!

cost ~ software mix + agency + agency size + data volume might be pretty interesting, though…

Okay, okay.

What this means is that a lot of weight will be carried by the best practices section.

And that means we should devote some thought to how it should look!

While we can mine the literature, we need focus.

Best practices are always best for some purpose.

What do we want to emphasize, in what proportion?

• Security?
• Cost?
• Speed?

The point of data is to let us know stuff. And know that we know!

Fundamentally, collections of data must:

• enable valid inferences regarding their domain
• even inferences not originally envisioned by the designer
• in such a way that those inferences are justifiable.

If you don't have this, all other considerations fall by the wayside.

As I said, this is going to be largely descriptive.

And we hope to tell a lot of our story graphically.

For instance, let's consider the questions about data volume and growth by DBMS.

We might see several things…

Questions?