Indices Long-term trends in the abundance of species are often communicated as indices. The abundance at some reference point in time is set at 100%. Sometimes this point in time is a range of years, often it is a single year. The reference point in time is something meaningful for the data: some special date (e.g. new legislation) or the start of the monitoring campaign. Although, seldom used, one could use the last year in the data as well.
When we analyse a mixed models, the question often arises whether a covariate should be used as a random effect or as a fixed effects. Let’s assume a simple design. Two types of fertilizer are tested on a number of fields (\(n_s\)). Each field is split in two and the fertilizers are assigned at random to these halves while making sure that each field has both treatments. From a conceptual point of view, we are only interested in the effect of the fertilizers.
inlabru is an R package which builds on top of the INLA package. I had the opportunity to take a workshop on it during the International Statistical Ecology Workshop ISEC2018 in St Andrews. This was a five day workshop condensed into a single day, hence the pace was very high. It gave us a good overview of the possibilities of inlabru but no time to try it on our own data.
Hillie Waning Vos posted a message in the Facebook group “vleermuizen” showing how she created some bat silhouettes at 1:1 scale. This is useful to illustrate the true size of bats to the public. She took a simple template and re-sized it to match the wingspan of the different species. Several people liked the idea. Some asked for the template or suggested to use a more anatomically correct template. E.
One of the reason why I often use INLA is because it allows for correlated random effects. In this blog post, I will handle random effect with temporal autocorrelation. INLA has several options for this. There are two major types of model, the first handles discrete time step, the latter continuous time steps. Dummy data set This blog post was inspired by a post on the R-Sig-Mixed models mailing list.
Recently, I got a question on a mixed model with highly correlated random slopes. I requested a copy of the data because it is much easier to diagnose the problem when you have the actual data. The data owner gave permission to use an anonymised version of the data for this blog post. In this blog post, I will discuss how I’d tackle this problem. Data exploration Every data analysis should start with some data exploration.
Introduction The default way to download recordings from a Peersonic bat detector is to connect the detector via a USB cable to the computer and copy them to the computer. The file transfer rate is quite low. In case you have to copy 100 to 200 files, this is OKish. But copying a full SD card takes half a day. In case you have a set-up with multiple detectors, this becomes a nightmare.
Git Target audience for this workflow Basic workflow Use case Principle Branching workflow with pull requests Use case Principle Branch Pull request Conflicts Flowcharts Rules for collaboration Exceptions Git For those how don’t know git, it is a free and open source distributed version control system designed to handle everything from small to very large projects with speed and efficiency. I use git daily, including for this blog.
Categorical variable Discrete variable Intro Fit with lme4 Fit with INLA Continuous variable Conclusion One of the questions to answer when using mixed models is whether to use a variable as a fixed effect or as a random effect. Sometimes it makes sense to use a variable both as fixed and random effect. In this post I will try to make clear in which cases it can make sense and what are the benefits of doing so.
I’ve reviewed the standard version the Peersonic RPA2 bat-detector in an earlier blog post. Today I will take about the RPA2 in a waterproof IP67 enclosure and my experiences on using it as on autonomous bat-detector. The differences The most obvious difference between both versions is their enclosure. IP 67 stands for dust tight and water tight under immersion up to 1 m depth. As you can see on fig. 1, the display and all controls are inside the enclosure.