Saturday, February 21, 2009

Making Figures

I have put in my 12 hours for the day, and my brain is whupped. Yes, whupped. But I can't settle down for the night, so I'm writing to clear my head.

I am currently writing the design section of my thesis. I thought this thesis would get easier after the painful introduction and background sections, and I thought the design section would only consist of pulling together the calculations I've already done. However, I've been here a year and a half, so some of the early calculations for my thesis were over a year ago. In looking back at my early notes in my lab notebook, all I can say is - what the hell? Sometimes, especially in my sketches, I can't make heads or tails of what I was thinking. Hey, didn't become an engineer because I'm good at art...

In the Excel sheets and FEA simulations I did, I didn't clearly label what I was doing. I thought I was clear, you know, "Study 1 - Thermal 2 - Run 4" nicely and neatly numbered. But which one was the RIGHT one? And the numbers are all there in Excel, and some numbers are bolded. But WHY did I bold those?

I always am sort of in a rush when I do analysis. Isn't every grad student? You just want to get the answer so you can go take data and turn out some results. I always intended to go back and clean up those figures, annotate that Excel sheet. I never thought it would be months before I revisited the issue, and with my poor memory it's only weeks until I forget what I was doing.

Ack. Only myself to blame.

So what I've had to do is basically re-do all my calculations. This has had two blessings: one, since January I have been taking AMAZING notes in my new lab notebook (I ran out of space in the old one). Two, I have discovered a few errors I made. Not large errors, thank goodness, but small things I should have tidied up. Assumptions I should have checked even though I knew they were good assumptions, for instance. For my thesis, I want precisely no errors, and no room for my work to be questioned.

So I've been doing figures. All day. Carefully and well formatted. You know what I mean (or you should, if you ever give Power Point presentations). All titles in the same font. Nice big font, of course. All data with a clear legend. Concise and helpful titles. Plot lines thick enough to see, in different colors for clarity, but also one dotted and one solid so if the paper's in black and white you can still read the chart. Always use the one consistent kind of plot line for theoretical results, and a different consistent kind of plot line for experimental results. Axes appropriately sized and divided, and consistent for charts of the same type.

And you know what the most disheartening thing is? If you do a figure correctly, it looks obvious. It looks simple. It looks like it took you no work. When in fact, it took you four hours to get it to look that good.

It's exhausting. But, lookie here. This is what I had, from nine months ago:

Ah, yes. Node 22837, of course, of Thermal Run 2.

Here is what I have now, with small errors corrected (note the different slopes of the lines now), simulation re-done (took me SIX times to get it to converge, sweet rainbows...), and graph nicely formatted:

Yay me. And you bet your buck tooth that I have three pages in my lab notebook documenting how I went about it this time.


  1. One thing I've had drilled into me- label EVERYTHING. write down EVERYTHING (even if it's admitting to screwing up). Two white powders are pretty indistinguishable after a couple of weeks at the back of the fridge and doing an NMR to work out what is what is a pain.

  2. You go girl! The revised one looks beautiful.

  3. Yay graphs! What are you using to plot? I'm a huge fan of R- I just write up quick little script files for specific type-graphs so when I need to plot an updated version it's blazingly fast....

    There should be a "sexy graph" club....