I’ve recently completed a three day weekend course that involved three groups of students forming proper hypothesis, designing a sampling program based on three choices of sampling methods (belt transect, point quadrat and quadrat), analysing our results and presenting to the rest of the class. Our final assignment is to write a scientific paper regarding.
However, the professor is looking for the following (sent to me by the TA when I asked for clarification):
‘As I understand the instructions you will need two hypothesis. One hypothesis relates to forest community structure that led to specific predictions you made about the density of trees in the disturbed versus undisturbed areas. The second hypothesis relates the differences in the sampling technique that leads you to specific predictions about which technique you think are better or more precise.’
Two hypothesis mean two predictions, two methods, two analysis, two discussions, etc., no? How do I go about writing a paper using two hypothesis? Would it be ‘Hypothesis No. 1, Hypothesis No. 2, Prediction No 1, Prediction No. 2, Method No 1, Method No. 2’…and on? I can’t find any help on Google on writing a paper in this manner.
(BTW, I was sort of thrown for a loop with this whole course as 75% of it was on statistics and statistical analysis, of which I have NEVER had any experience with. Talk about a huge learning curve!)
No, you don’t need to write two method, results and discussion sections. Just break them down into subsections. Most scientific papers have multiple hypotheses to address.
I wouldn’t necessarily duplicate every section (ETA: subsections are fine ) of your paper.
Look at it this way:
It’s possible that the differences you see in your different samples are due to differences in the basic structure of the forest (I could be more specific - but I don’t know what you were measuring - canopy cover? diversity of vegetation? undergrowth?).
It’s also possible that the different samples are exactly the same, but you used different methods to measure them. The technique you choose to make your measurement should affect your results, that’s why some techniques are preferred over others in some situations.
To write this, I can’t see that you would possibly include two separate introductions, methods, results, discussions. What I would really want to see are in the results, a side-by-side comparison of the three techniques you’ve used. In the discussion, I would want to see some acknowledgement of the fact that both sampling technique and fundamental differences between different locations in the forest could explain disparate results. A discussion about which hypothesis you prefer to explain your results and what evidence would favor one hypothesis over another would be necessary. Also, a search of the literature will reveal papers discussing the selection of sampling methods and the effects choice of method will have on your results. Include references to what is known about your sampling techniques.
YMMV, I am not your instructor, so you should probably get clarification from them if you don’t understand, but that’s what would make sense to me.
To clarify, we were comparing a recently burned forest to a mature forest. Our group hypothesis was that the ratio of Lodgepole Pine to Aspen in a mature forest would be different then the ratio of Lodgepole Pine to Aspen in a recently burned forest. Our prediction was that the mature forest would have more Lodgepole as compared to Aspen as compared to the burned forest. Unfortunately for us, we were either really correct, or we didn’t do a big enough sample, and in the mature forest 75% of the plots had zero Aspen, which meant we couldn’t do an analysis. (We were able to get around it in the end after staying up until 1 am the day before drinking beer and tossing around ideas.)
Anyhow, your advice clarifies it quite a bit - actually I do recall him saying something similar to ‘In the discussion, I would want to see some acknowledgement of the fact that both sampling technique and fundamental differences between different locations in the forest could explain disparate results’, and you’ve just reminded me, so thanks. As for getting in contact with the professor, I tried, and he’s conveniently on vacation until the day the paper is due!
Aha, I get it. How big were your belt and quadrats, if you don’t mind my asking?
Not able to do the analysis? I’d be suspicious - you should still be able to analyze the data with a number of zeros. Or are you required to use a specific statistical method? I can see how you’d want to avoid assumptions of normality in this situation.
That’s always the most convenient time to go on vacation!
Our belts were 2m x 10m with randomized locations, the quadrats were 100m square, again randomized.
We used the two-sample t-test on our data. Because we had zero’s, they couldn’t be used as the denominator, making it impossible to analyze. Our options were that or one-tailed t-test.
It sounded to me that your hypothesis was to calculate the ratio of the counts of Aspen vs lodgepole. A t-test, either one-sided or two-sided, tests means. It is not perform all that well at testing ratios of things because variances of ratios are tricky. A two-sided t-test infers whether the mean of two things are equal (under the null hypothesis), versus a difference (alternative hypothesis). That is a different hypothesis than a ratio hypothesis.
Just because you have zero counts, doesn’t mean that you can’t perform a test. You could set up a contingency table and test an odds ratio by using a chi-sq test.
The sampling unit here is quadrats I think.
Enumerate the following counts:
Mature Forest quadrats with Aspen
Mature Forest quadrats with lodgepole
Burned Forest quadrats with Aspen
Burned Forest quadrats with lodgepole
Apply chi-square if your sample size is big enough. Otherwise use Fisher’s Exact test.
The following will be of no help to you whatsoever, since you’re talking about something different:
Back in the bad old days when I was majoring in biology, every experiment we did had, by default, at least two hypotheses. The first one was the real hypothesis, and the second was the “null” hypothesis.
Hypothesis: under conditions X, Y will happen
Null hypothesis: under conditions X, Y will not happen
That way, you get to feel happy about one hypothesis or the other being confirmed. Failure is always an option.
) of your paper.