Friday, January 4, 2013

Coronagraphs and SVUSD.

Arthur Eddington's photograph
of the May 29, 1919 solar eclipse.
available at http://tinyurl.com/n46h3z
The photograph at the right is probably the most important one taken in the 20th Century. It was taken of an eclipse of the Sun by Arthur Eddington, on May 29, 1919, from the island of Príncipe in the Gulf of Guinea, off the west coast of Africa.  In the upper right portion of the Sun's corona, you can make out a small dot, which is the Hyades star cluster.

The reason it was so important, was not because the position of the Hyades star cluster was wrong, although that was definitely the case. The Hyades star cluster has "moved" in this photograph when compared with all of the other stars in the sky -- but the problem was that it moved too much.  

Isaac Newton had predicted hundreds of years before that gravity could bend light.  He had even predicted how much the light would bend.  But in this rare instance, Isaac Newton was wrong -- the light from the star cluster had moved twice as much as he had expected. 

But someone else's theory was right -- a former Swiss patent clerk by the name of Albert Einstein.  The apparent movement of the Hyades star cluster was exactly the amount that Einstein's theory predicted when light traveled near the warp in spacetime induced by the Sun's gravitational field. 

OK, nice story, what's the point, John? 

This post isn't about relativity, or even really astronomy. It's about how this experiment was possible in the first place. The Moon, in an eclipse, functions as a natural coronagraph. Like any coronagraph, the moon allows the viewer to see the light from the corona of the sun (and the light from the Hyades star cluster), which would otherwise be hidden in the Sun's bright glare. Without the moon being in this position, measuring the degree of the apparent movement of the light (and the proof of general relativity) would have been well-nigh impossible. 

Why am I talking about this today? Well, over the past week or two, I've been looking at the test scores for Sonoma Valley's elementary schools, trying to develop a picture of what's taking place from the data.  In my last post, I used a very limited set of data to illustrate a point -- the overall rankings of the schools measured by API scores, and then the scores of a particular subgroup across different schools over time.  Those two pieces of data point in different directions.  

In talking over some of the comments made by some very intelligent people who reviewed the post on Facebook, it became clear that they were aware that I was using something akin to a "social science coronograph" to block out a particular piece of data, which was so bright and glaring that it would make it impossible to view what I'm really interested in. 
Ethnicity of Students, Hispanic or Latino
Sonoma Valley Public Schools
"Education Statistics of California," 
Google Public Data Explorer
I deliberately blocked out that piece of data, because, like the Sun itself in Arthur Eddington's experiment, it obscures more than it reveals.  But I don't want anyone to think I'm hiding the football here -- far from it -- it's just that this particular piece of data drowns out the signal.  

That piece of data is the graph at the right. It's the percentage of the population at each elementary school that is Hispanic or Latino. This graph is essentially the reverse of the API score decline graph I posted yesterday.  Google's setup isn't letting me graph ESL students, but the graph is essentially similar when you review it at the State of California's site. This demographic change happens to make the API scores vary wildly, which was one of the big reasons I picked a narrow subgroup when evaluating the schools over time. 

As you can see from the data, Sassarini, where I'm thinking about sending my daughter, is about 2/3rds Hispanic or Latino as of 2011. That's not a drawback from our family's perspective. My wife and I have been making an effort to ensure our daughters have the opportunity to learn Spanish, and perhaps the great overlooked point in learning a language is that you need to listen to it.  The fact that our daughter would thus be exposed to a great deal of Spanish on the playground is, frankly, a positive. I am, however, aware that there are parents that disagree.

But on the question of whether a school is improving or worsening in its job of teaching children like my daughter, the ethnic makeup of the school is irrelevant. And that's why, in making that determination, I pulled out a virtual coronagraph and blocked the glare from this graph ... it's not something that is inappropriate for consideration -- it's just that I don't think it reveals anything about the statistical performance at a particular school, and that's the thing I'm interested in measuring at this point.

There's a second issue that some of the comments were pointing out.  This was the fact that I left off the charter schools on these graphs, and that I didn't point out the scores of some of the specialty programs that exist in the school district, such as the dual immersion program at Flowery.  

The second point is more straightforward to address than the first -- the dual immersion students appear to be included in the raw Flowery numbers.  

The first point is nuanced.  The charters have unusual characteristics -- for instance, Woodland Star is a Waldorf school, and trying to measure a Waldorf school via something like an API score would be pointless and ultimately deceptive. If you want to understand a Waldorf school, you're going to need to meet the families involved.  I can tell you from experience that I am impressed with the Woodland Star families.  So, I would toss out the API score as a measure completely when evaluating Woodland Star. 

Sonoma Charter has a similar problem, but in the opposite direction. Sonoma Charter's Hispanic/Latino numbers are consistently around 15%, which is generally about half that of Prestwood.  When that issue is adjusted for, the remaining 16 point differential between the students at the two schools is modest at best and is probably irrelevant. But as a parent who takes Spanish education seriously for my kids, my evaluation of the quality of the school is impacted by the skewed characteristics of the population of the student body.  

Finally, on the point of whether test scores are a good thing for parents to look at -- I think they reveal a lot of things, many of which were not necessarily intended by the creators of the test.  Teaching to a test is always a concern, and a well rounded education is a means to an end -- the ability to function as an informed citizen in modern society.  Cram courses to learn how to fill out little bubbles on a Scantron can improve test scores but do little in terms of educating children.  

However, there are a lot of parents in Sonoma whose children will probably go on to national-level (if not international level) institutions for postsecondary education, and dealing with the reality of standardized testing is a necessary part of preparation for navigating those institutions. Scores should be taken with a grain of salt (or ignored completely when evaluating specific programs), but ignorance of the nature of the testing regime is probably more dangerous for parents and children in the long run.