Wednesday, April 30, 2014

Formal Learning: Data, Big Data and Statistics.


The first challenge we can call an input problem.  Developmental disabilities are defined statistically for the most part.  In statistics

This will start what I hope will be a series on formal learning in individualized services.  Because I sometimes read, and all the writing these days is about big data (100.5%.)  There are corners of the system (almost vacant of service providers and probably of families) that hope that collecting outcomes data will lead to better services.  I think that hope should live on, but that people understand that the benefit will be much less direct than in other sections of the economy.

To give a quick overview of "big data" and its benefits, now that so much of what we interact with generates information and the capacity to store and analyze it so much vaster than it had been, that humans have grown much much abler to discern patterns that had escaped us in the past.  The opportunity to make change comes when we are able to see those patterns in context.  Using statistics, we can find what factors affect the patterns we are concerned with the most.  The term of art for the factor most relatable to changing a pattern is "the big coefficient."  Terms of art in statistics are still pretty arcane and prosaic.

From my spot on the spectrum, individualized services ought to include the search for, identification and exploitation of patterns along with respect, protection and kindness.  And math, particularly statistics, are the handiest tools we have with which to do that.  And the rest of this series, if and when it emerges, will be about why I think professional caregivers should do math.  But there are reasons to question whether big data can have the same impact in this field that it already has in medicine, marketing, science, politics or engineering.

One reason why big data will have trouble helping us help the people we serve might be called an input problem.  Just to clarify nomenclature, to the left is what is called a "normal distribution."  Of any given naturally occurring trait, there is a central tendency where any random individual is most likely to fall.  That is called the mean and can be pictured as a line through the highest part of the curve.  Where the curve flattens out to the left and right is called "the tails."  I bring this up just because the word "tails" can give either the sense of disparaging or adorable and I wouldn't want to be thought to mean either.

But most disabilities are defined at least in part by a trait being found in an individual to occur in the tail of the distribution.  You can imagine a stone dropped in still water.  Where the stone strikes, you get the most information and further along in the eddy you get less.  Not only are the people we serve rare, but it is easier using statistics to learn about commoner individuals than about rarer ones.  Which is just to say that coca-cola will still know more about refreshment-seekers than DDS will about people with developmental disabilities even after the latter starts really trying.

The other problem we can call an output problem.  When KFC wants people to eat more chicken, it is easy to find factors that correlate with the sought behavior.  If the state wants fewer people to be poor, it is relatively easy to use large data sets to figure out which factors have the profoundest impact (largest coefficient) on poverty and proliferate them.  But the Lanterman Act and those of us who serve it, wants people to live the lives they choose, not to behave according to a standard.  And that makes it much harder to find the large coefficient independent variables.

So now I hope to write upcoming posts about why measurement and math belong in the complex of tools states and their agents use in pursuit of our mission.  But I hope this post set some boundaries on how much we can hope to accomplish this way.

7 comments:

paul said...

Doug:

What does that data that forms your “patterns” look like, and how do we reliably collect that data? Can we have a discussion about patterns, without first clearly defining what it is that forms our patterns? Furthermore, the “exploitation” of patterns based data that is unreliable, wrong, or based on a prejudicial preconceived reality can do massive harm. Arguably, past analysis of patterns based on improper data and prejudices is what gave us Developmental Centers.

Perhaps the best illustration of my point is your statement : “Lanterman Act and those of us who serve it, wants people to live the lives they choose.” No, it does not. The Lanterman Act is intended to provide supports that permits a person to approximate the pattern of everyday living available to people without disabilities of the same age. What does that mean?!!! And, how do we collect data about that?

Primarily, I wish emphasize your point about how difficult it is to get reliable data. Secondly, I want to point out how easily our personal desires to define what we do can mess with things. I don’t know you Doug, but anyone my age who approximates my pattern of everyday living will discover very little choice in his or her life.

The NCI survey tells us that 1 out of every 10 people with disabilities said they have a paid job in the community and that 90% of that 10% are happy with that job. But, it tells us nothing about why a person likes his or her job (it has GREAT vending machines!!), whether the job is providing anything other than a wage, nor does it tell us anything about the job (perhaps in part the service being provided by the system).

Give me one tree, and I will visualize a forest

paul said...

Also – are you posts about this subject going to be a philosophical/academic discussion, or something more applicable. E.g. Is this a world where we have a 10 billion dollar budget and everyone is nice as Stanley, or do we have a 5 billion dollar budget paying with about 10% of us as nice as Stanley.

Doug Pascover said...

Great questions.

Regarding the patterns, that's what I'm hoping to establish. How useful will it be to aggregate data in the hope of providing better service to one individual? I think it will not be useless but I also don't think we'll be able to develop a how-to guide that way.

I also want to hold up your third paragraph, although I don't take it as on-topic: But something I've realized in my work is that the "person-centeredness" which is now the standard goal of DD services is highly anomalous to a life similar to the non-disabled peers.

But I do think two valuable things will be possible with aggregated data. We will become much more able to find interventions and agencies that do not work or that are much less effective than alternatives. And we might be able to guess at interventions and agencies that will help a certain person with a certain goal.

That's where looking for patterns can help. There are currently many agencies that fairly or unfairly have the reputation for not helping people get jobs and those agencies are about as full of people as the ones that have the reputation for helping people work. Some data on this might help individuals elect effective services and help eliminate wasted time, money and life energy.

In one of the upcoming posts I plan (whether I get to it or not) will be about how math (loosely) can be used with one individual to find patterns that better enable staff to prevent crises or seek better results. That's something I'm developing a training on.

And I partly concede your point about living lives of choice. As I recall it (and it wouldn't be a lot of work to quote but here among friends I'll just go from memory) the purpose is given as living lives of their own choosing that reflect those of non-disabled peers. I don't think the crafters of The Lanterman Act saw the conflict you and I see. But it's important to remember that when that law was enacted, individuals with developmental disabilities had much less choice than their non-disabled peers. But living a person-centered life is radically different from one that reflects that of non-disabled peers.

Doug Pascover said...

Paul, to your second comment, most of the posts that have ever appeared here are predicated on the idea that we can only make changes that can be made within existing resources. I see no value in trying to imagine a $10B system where nobody was crankier than Stan.

My defining trauma in policy-making and management was the System Reform effort during the Davis administration where I watched my colleagues and their paid advocates try to claim that it takes a college degree to hoist a human.

stanley seigler said...

nice to hear from ya'll ... i'm a couple months behind as these posts did not show up the browser (whatever that is) i use ... just happened to click on an old browser file ... anyhow i continue beating my dead horse, IPPIDT!

not sure just what's being said re the big pic (philosophy, i guess) ... but some specific comments

paul sez. "The Lanterman Act (theACT) is intended to provide supports that permits a person to approximate the pattern of everyday living available to people without disabilities of the same age. What does that mean?!!!"

dunno but what theACT sez, the state shall fund support/services as determined by the IPP. the IPP needs shall be determined by an interdisciplinary team (IDT)

'What does that data that forms your “patterns” look like, and how do we reliably collect that data?' [paul]

and what conclusiond can be drawn from the patterns...will it tell us if IPP goals are realistic and being met or adjusted?

stanley seigler said...

nice to hear from ya'll ... i'm a couple months behind as these posts did not show up the browser (whatever that is) i use ... just happened to click on an old browser file ... anyhow i continue beating my dead horse, IPPIDT!

not sure just what's being said re the big pic (philosophy, i guess) ... but some specific comments

paul sez. "The Lanterman Act (theACT) is intended to provide supports that permits a person to approximate the pattern of everyday living available to people without disabilities of the same age. What does that mean?!!!"

dunno but what theACT sez, the state shall fund support/services as determined by the IPP. the IPP needs shall be determined by an interdisciplinary team (IDT)

'What does that data that forms your “patterns” look like, and how do we reliably collect that data?' [paul]

and what conclusiond can be drawn from the patterns...will it tell us if IPP goals are realistic and being met or adjusted?

Karen G. said...

So basically, Paul, the boundaries the executive director of imagine is trying to introduce is the concept of "compartmentalization."

This means not saying "Amen" to me within six minutes of my reply to commenter on a totally unrelated topic.