I'm typing out my final lab report for my independent study into Epidemiology, and I've run into a slight snag.
My experiment consisted of a couple weeks of surveying a group of students and comparing the amount of sleep they were getting and the stress they were under to the amount of bacterial growth that occurred when we plated a sample from them.
The first three surveys show a very high correlation between stress and bacterial growth.
I'm talking about an almost perfect logarithmic fit to the data.
The final two surveys, however, have data all over the board. People with no stress registering a hundred colonies, and people with extreme stress and no sleep with literally no growth on the plates -- and that isn't right under any condition: we all have bacteria; it should be showing up.
I examined the problem, and, after trying a few tests, came to the conclusion that the nutrient agar I had grown the final two samples on had been badly made. Heck, I even tried a swab from the school toilets on it and nothing grew. This may have been because of poor autoclaving or even excessive heating during the making of the agar.
So, when I'm typing up my analysis, can I refer only to the first three surveys and generally omit the final two, and then explain why in my conclusion? Or is that innapropriate?
I believe I do have to include all the surveys in the data, but if a set of data is so obviously off-kilter, why should I include it in my analysis?
Posted by King of Men (Member # 6684) on :
You are certainly permitted to disregard data if you ahve reason to think it is bad, but this should not be relegated to the conclusion. It should be in the section where you explain how many samples you took, and you should also outline in some detail the tests you made that make you believe the cultures were bad.
Posted by Shigosei (Member # 3831) on :
I'd recommend that you include the data and discuss the fact that you believe that the nutrient agar is faulty. You should definitely mention that you tried to grow bacteria from a toilet.
You can, however, still make the conclusion that there is a correlation, even though the last two data sets don't show it. Once you've explained why the sets are faulty, you can go ahead and discuss what the first three sets demonstrate. If you aggregate data from multiple surveys, I'd go ahead and use only the first three sets. You should report the data; you don't need to necessarily include it in your analysis.
Posted by aspectre (Member # 2222) on :
Besides, that's how penicillin was discovered: Fleming noticed that certain culture plates didn't grow bacteria, and included that fact and his subsequent investigation in his published report on his experiment. Years later, someone else noticed Fleming's paper, and did some further research deliberately cultivating Fleming's culture killer. Out of which was eventually developed a new class of antibiotics. Without Fleming's willingness to point out his own "mistake", millions upon millions of people would have died from now easily treatable infections.
Always remember serendipity: The truly INTERESTING parts of your experiments may have nothing to do with what you were trying to accomplish.
[ May 18, 2005, 03:57 AM: Message edited by: aspectre ]
Posted by HollowEarth (Member # 2586) on :
It looks (and is) much better if you can justify throwing the data out. If you had a protocol for use of the agar, it might tell you how to do this. Showing statistically that these are outliers works too. If you had control plates that are off this is probably good enough.
Posted by El JT de Spang (Member # 7742) on :
I was always taught that you were not to be held responsible for your screw-ups, as far as lab reports are concerned. The important thing is whether you can explain what you screwed up. The "why" is always the most important thing.
Mention them as part of your procedure, but explain why they had no growth, and thus why the data they produced was thrown out and not used in the results.