One of the exercises in Ecological Statistics fits a species area curve to data from the Galapogos Islands. I ask students to compare the estimated power coefficient \(z\) with the value 0.27 reported in a meta-analysis by Drakare et al. 2006. But Drakare et al. only reported the mean, not the standard error.1
To do the comparison properly, I need to know how uncertain the average \(z\) is. I looked through the paper carefully, but couldn’t find a standard error anywhere!
Many years ago I started making videos of my lectures for my ecological statistics course. I still use those early videos, now on YouTube, but a recent comment on the checking assumptions video made me realize that I somehow misplaced the code that I used to create the figures in the powerpoints of the early videos. 1
Rather than try to reproduce the old code in a non-public way, I thought I would redo it here so it is easy to share.
I teach my students to check the assumptions of their models by making various diagnostic plots of residuals. One of the niftiest is the scale-location plot, which is useful for diagnosing changes in variance across the range of the model. If all’s well, a smooth line on that plot is flat. But how flat is flat?
The problem is that real data is never “flat” even if all the assumptions of a model are met.
One of the exercises I make Ecological Statistics students do is work backwards from a given value of the dependent variable in a regression to the corresponding values of the independent variable. I usually just get them to eyeball it from a table of predictions. But today a student suggested the “terms” type of prediction might be the answer. Unfortunately I’ve never tried to figure out what type = “terms” actually does1.
5 out of 10 ecology textbooks on my shelves make this distinction: geometric models are for populations with discrete pulses of births, while exponential models are for populations with continuous births. This zombie idea needs to die. It is both wrong and enourmously confusing to students.
I set up 4 10 gallon aquaria in the teaching lab this semester. The goal was to observe the nitrogen cycle as the tanks settled in. I also needed an excuse to try out the new Hach environmental water testing kit that we ordered. The data are on figshare.
A colleague wrote:
Hi Drew,
I’m trying to make a really simple (I think) plot in R, but am not sure how to do it. I want to make the attached, where the size of each “bubble” in each grid location depends on a single raw data point (% cover of a type of grass). Can you point me in the right direction?
This is the goal.
Really. I have no concept of what my students' lives are like, and this is something I have to constantly remind myself. I cannot make assumptions about their situations based on my own experience.
I've been teaching students `ggplot2` for graphics exclusively for a year or more now. One issue that seems to throw students is the specification of different data sets for some layers. Part of the confusion seems to arise from reversing the order of arguments between `ggplot()` and `geom_*()`. I'm trying to combat this by always naming my arguments. Is this a good idea?
Should ecologists use sum-to-zero contrasts?