Mistakes Were Made

Recently, I've been compiling and analyzing my data, and I noticed my trends were not all clean-cut increasing. In fact, population density actually declined in several of the groups exposed to our "medium" amount of oil, or 3 drops. Below are graphs roughly displaying the ratio of the experimental group's population density average to that of their control groups (CH meaning Chaetoceros and the number being the first or second trial and TH meaning Thalassiosira.) The x-axis is the oil drop amount and the y axis is the ratio to the standard. The data point at (0,1) represents the standard itself.

As you can see, the trends aren't all nice and pretty (and please ignore the typo in the second graph title). So I emailed Dr. Edward J. Buskey, the director of  the Gulf of Mexico Research Initiative DROPPS Consortium (whom I had previously emailed about the mysterious ghost and stretched out cells), asking if he had seen similar patterns in previous experiments from his lab and if not, if he believed my results could be due to some outside force (altitude, light level, etc.). His response was as follows:

"Your exposure levels are much higher than we have used in our experiments – your highest levels of 1660 – 2500 ppm (1.6 to 2.5 ppt) are much higher than those used in any scientific study on the effects of oil on phytoplankton that I am familiar with. I would not expect exposure to such high levels of oil to stimulate their growth."

This I considered somewhat weird. Though this amount of oil was nearly impossible to occur in a real spill, why had no researcher ever wondered what the upper limits of oil a cell could metabolize was? Was science so limited to real-world situations that people stopped wondering "Why not?" when faced with testing hypothetical circumstances? He then followed up with this:

"Counting cells can be very challenging – you need to make sure your culture is very well mixed before you sample it since diatoms will tend to sink to the bottom of a culture; you also need to make sure you count only live cells, and that they are not dead or dying cells. And you must count enough cells to eliminate potential counting errors. Fro example, if you only count 4 cells in a sample, your 95% confidence interval (your chance of making an accurate count) is +/- 100% !!  If you count 100 cells it is +/- 20%.

You need to count a lot of cells in each of your experimental treatments  to avoid chance counting errors."

But I hadn't been shaking every bottle we tested to ensure it was "well-mixed," only a select few which I thought had a large proportion of cells settled at the bottom (as in, I'd shake every bottle in one trial, but not during the next trial. I wasn't just shaking up random bottles in each individual trial). And I hadn't been differentiating between dead or living cells (it's hard to do that when A.) a cell is non motile and B.) you're basing your counts off pictures). AND most of my sample pictures usually had below 100 cells in them. Sheesh. Looks like my error margin is going to be as big as the Marina Trench is deep. 

Signing off till next time, this is Erin Butcher