More about work

Time for status update #2. Basically, since I got back, we've been working double-time, and this is the first night I've even had enough break to write something. My trip back to Bhorugram from Delhi was in the company of our computer guru from Hyderabad, with whom I spent three days trying to install and configure our new server and four new client computers to match it. You'd think this would be easy, but when you're six hours from any serious tech support, nothing is easy. Everything in India is done with pirated OS and pirated software, and hardware is patched together as custom jobs. Net result: we had a hard drive that was wired wrongly into the machine and multiple machines riddled with viruses. With heroic efforts (and not celebrating Holi in the traditional fashion of getting seriously high/drunk), we got it marginally working. Getting it production-level working is still in progress.

That left me with two days before the big bosses themselves, Drs. Reddy and Agarwal, came to down. So, I decided to implement some of that quality checking, using two time-honored statistical techniques: the convenience sample and the n-of-1 study. Namely, we printed out our data concerning the anganwadi centre located in Bhorugram village, traipsed down there, and started hassling the workers with our questions. Once we'd gone through everything she had written down in her register, we spent two half-days marching around the village (with plenty of people staring at the mini-parade), wandering into houses and investigating vaccination cards, numbers of children, pregnancy dates, and so on. This was clearly quite an event, even for a town that sees plenty of strangers -- we had at least two local children following us at all times, and any entry into a house meant a small crowd of neighbors outside.

Unfortunately, the fieldwork revealed about what I was afraid of -- even with records on over 250,000 people in the database, we are missing a pile of information. We had records on 20 births in 2008. In a pile of papers not a kilometer from our computers, we found records of another 20. We knew of 6 pregnant women; the workers knew of 5 more. Mind you, these are the same workers who fill out the paperwork on which the database is built, and on whom we depend for our monthly updates.

So, what do you do when you've spent a whole lot of time doing a cross-sectional survey of 233 villages, only to find out that it looks like you might be missing a lot of data? That was the topic of discussion with the "attendings". Thankfully, the answer will not be "start the whole dang thing over". It will, however, be "Make some village women do more work." Basically, since we know they've got much of what we need hiding in their registers, we hand them a form and tell them "fill this out, we'll want it back in a week". Not necessarily what I'd have chosen to do (I don't trust their reports, as the forms are printed in English, which most of them don't read), but I'm only the local manager around here. It'll do for a short-term fix, which is all we have time for before I pack up and go in another 1.3 months.


The big question for me right now is how to maximize that 1.3 months. I have a lot of potential focus areas, which I'll try to describe in a future post.
Meanwhile, while I was gone, one of my staff flat-out quit and another accepted a position to begin in two weeks. On the bright side, I got added a really smart statistician. So, I need to figure out how to use these new resources, and how to design something that'll survive this highly unstable environment. "Challenge" is putting this mildly.

Luckily, at least as far as blog entries go, I've got enough pictures and pre-written general posts to keep you entertained for a few weeks. :-)