Necessary work productivity has trumped blog productivity this week and it was going great until today. I was moving on a project that I have been really excited about, but has been on the back-burner for the past several months. Instead of just admitting that this project has taken the position next to the tea kettle that is never used in the summer months and writing my more senior collaborators on the project to this effect, I continued to delude myself into thinking that I was going to get it done. Of course, this meant long periods with little work and great disappointment or other periods where I would get such minimal work done on the project that it was laughable that I would even send it to the other collaborators.
Anyway, this week, I made huge strides forward towards accomplishing the basic analysis of this project. I either obtained or created all of the variables necessary to replicate a previous analysis, completed all of the data management tasks to line up the proper hierarchical structure and even conquered Stata's best attempts to get me to give up on using Sean Reardon's Stata interface for HLM[1]. Anyway, I got all of the analysis complete and then I tried to figure out how to run spatial analysis in Stata. Needless to say, it doesn't come with a built-in package and my old process has been to use a separate program, GeoDa, to do all of it.
I am, however, quickly getting tired of having to use three or four separate software packages to work on a single problem. Therefore, I again invested the time in attempting to figure out how to use a spatial analysis package built for Stata. That is where the "things were productive until today" part enters. I could not, for the life of me, figure out why I could not get this to work. First of all, the package requires four separate utilities just to get the data formatted properly for analysis and each step compounds any problems created in the previous steps. Anyway, I finally figured out that I had a single missing value in a single observation which was holding the whole thing up. Then, when I finally got the proper data form to get the program to run, it didn't agree with the values that I got in the other program, making me wonder why they come up with different results which means more time trying to figure out the shady underbelly of programming even though any programming knowledge I do have is self-taught meaning that it is both laborious and probably wrong meaning more potential delays in this project.
Therefore, following ash's excellent example to productively (hopefully) blowing off steam, I dedicate this entire day to hating a) Stata programs that are difficult to figure out; b) Stata programs that don't work and c) loss of productivity to said problems.
And now, I am sure that you are hating me now for writing such a boring post. G'day.
[1]For anyone who uses both Stata and HLM, it is worth the time investment to get this program set up and learn how to use it. No more switching back and forth between data management in Stata and analysis in HLM. Of course, I guess you could also just use SAS, but learning a whole new programming language is not an investment I want to make right now...
