Critique of Forest Patch Area and Butterflies Writeup
I have looked through a couple of students' writeups, and I have some suggestions about language.
As your primary guide, use the language in the Case Studies of the Sleuth.
Here are some comments on the language used in the sample one-pager on butterflies:
- Summary stats should include resistant measures in addition to mean and SD. Include median and IQR.
- Replace figure 1B with a figure showing the curved relationship on the original scale of X and Y.
- "lack of stability in variances" should be replaced with something like: When fitting a regression of Y on X, the error variance
appears to be non-constant.
- "Transforming...results in a better distribution". Better way to say this "Transformation of the data by taking the log of species and
the log of area and running the regression of log of species on log of area appears to resolve the non-constant variance
problem. In addition, the residuals of the regression have mean zero at each level of the fitted value of log(species), and appear
roughly normally distributed."
- Eliminate the language "normalizes the SDs of the groups". I'm not sure what this means. Again, refer to the language above. We aren't normalizing variability; we are transforming to achieve constant variance in Y across the values of X and to ensure that the residuals of the regression have mean zero at each X value and are normally distributed.
- Report the fitted line in the format I have asked in the HOMEWORK. The butterflies version is incorrect.
- If you report estimates, use hat notation using the Word Equation Editor.
- The writeup talks about a "multiplicative change". You should specify whether it is a "multiplicative increase" or a "multiplicative
decrease".
- Delete the part of the sentence "The p-value for the hypothesis Ho...means that". Capitalize the "I" in "if", and put at the end of
the sentence after the word "experiment" -- "(two-sided p-value=.002).
- Don't give the null hypothesis using "Ho: beta = ". Rather, write the alternative hypothesis in words, "We performed a hypothesis
test to test the claim that intelligence is linearly related to shoe size."
- Eliminate the sentence "Therefore there is sufficient evidence to reject the null hypothesis." Put this in the words of the problem.
- Get rid of the extra sum of squares F-test sentence. It should be replaced by a t-test of the claim that the rate of change of Y
with X is nonzero. (Of course you'll specify X and Y in words)
- Note that the student removed the 2 outliers and reran the regression. This would be followed by a comparison of slope
parameters and respective confidence intervals. Don't need to report the new regression itself unless it is considerably different
from the old. Also comparing R2 values is of little use here.
- In the conclusions, "demonstrate a significant relationship" should be replaced by "demonstrate multiplicative increases in the
number of species are associated with multiplicative increases in the median area."
- "Further conclusions are dangerous to make" is vague. Be specific about what conclusions can't be made from this model, and the
limits of the data in answering the research question. Suggest specific confounders that should be measured as well.
Last modified: Thu Feb 14 23:33:44 EST 2002