### Introduction

for my course Analysis of Spatio-Temporal Data in our MSc program Geoinformatics. Barry kindly answered:

Well, here you go!

### The course

The course has a a goal to get familiar with what spatio-temporal data is, and how you can analyse it, with emphasis on using R. Analysis methods focus on spatial statistical methods (point patterns, geostatistics, lattice data) and analysis of movement data. The course consists of lectures and exercise meetings; for exercise meetings students were supposed to hand in short assignments. The larger final assignment is individual, and is like a small, reproducible research paper and forms the basis for the final grade.

Instead of (or: in addition to) working with pre-cooked examples that use data from a particular package, loaded by data(dataset), I encouraged students from the beginning to look for datasets themselves, and import them in R. This way they learn, and learn from one another

• how datasets look like in the wild
• how they are imported in R
• the hoops they have to go through to get time or date variables right, e.g. by using as.POSIXct or strptime and
• troubles that coordinates may give, e.g. their projections or degrees-minutes-seconds notation.

as well as challenges related to applying methods:

• time series analysis does not work if you have very short time series
• periodicities are often absent in yearly data, but present in sub-yearly (daily/weekly/monthly)
• spatial correlation is hard to assess with a handful of observations
• not all point-referenced datasets are meaningful input for point pattern analysis, or for geostatistical analysis
• most point pattern software assumes Carthesian coordinates, but does not check for this

### Reproducing

I’ve asked students to hand in assignments as R markdown files so that I can not only see what they did, but also redo what they did. I’ve asked them specifically to

• work in a separate directory for each assignment, using lastname.assignmentNumber as directory name
• include the R markdown file, data sets, and the output (html or pdf)
• not set paths in the R markdown files
• not use absolute paths to data files
• not use install.packages in their R markdown file
• zip this directory and submit it to the moodle system

I’ve also demonstrated for the whole group how things work when I reproduce this, and showed them where things go wrong. For reproducing, I have to