Wednesday, January 7, 2009

Learning from other people's mistakes

Lessons learned from mistakes have always stuck around in my memories the longest. A lot have been mistakes I made, but seeing other people screw up is also a great source "what not to do"s. This is especially true when you have to work with the results of the fuck-ups.

I spent most of my summer at Research Organization trying to clean up the ugliest database ever because the person who made it was really, really incompetent (Seriously guys, if you're collecting data electronically you should never, ever let people enter whatever they want. Even [Especially?] if said people have an MD). I managed to clean it up well enough that we've already got 2 abstracts accepted for a meeting and there was talk of writing a manuscript. Along with these easily quantifiable achievement I also learned a lot of shit. I learned how to work with a particular statistical package that isn't SPSS, ethical data cleaning, and how to politely tell people that they're fucking nuts. However, the things I will remember forever and will hopefully never repeat in my own work include "Why yes, enter whatever you wish for the units" and "Hmm...contradictory information, nobody would ever enter that so why should we create measures against it?".

Over the break, I started and almost finished making data entry forms for a different study. They're nice, shiny and they are made to be as idiot proof as possible. But again other people screwing up is fucking with my work. This time it's not the kind of mistake that leaves me with lots of interesting work and amused frustration at incompetent people, it's the kind of fuck up that will most likely render 3 weeks of work a complete waste of time.

The gist of the problem: pretty science isn't always practical science.

The study is a clinical trial looking at a novel use for a very common medication. The scientific justification is so simple and sensible that even a person with just a little background on the disease (me) can understand it easily. The study is also well designed for the most part. However, the problem is that the protocol requires a super strict time frame that allows for very little time between an occurrence of the medical event in question and seeing a specialist. This makes some sense as it would make the science better. However, the time frame is so ridiculously ludicrous that at the rate the PI is enrolling patients we're probably going to have a case study rather than a clinical trial.

God, I hope that if (when?) I ever have to plan studies or even simple projects I'll have the foresight to be able to tell when my plans are fucking impossible.

P.S. If none if this makes sense or follows a logical progression, I blame anger and Jagermeister.

1 comment:

Eugenie said...

I totally get you on the data entry front. People who are chronically afraid of statistics should not be allowed to enter in data! Over the summer I had to fix unit fuckups, missing data fuckups, among other things. This time, the offender was a DVM. We should be able to slap these people!!