Description Statistical modelling comes with a range of mathematically convenient assumptions impai ...

---

Description Statistical modelling comes with a range of mathematically convenient assumptions impairing our ability to accurately model the world around us. One of those assumptions is that observations are independent of each other (i.e. assuming exam results are independent when we have one good teacher and one bad teacher). This is often not the case in humanities research, where researchers need to commonly overcome a problem called autocorrelation. We will look at the commonly used linear regression models and how we can simply account for the non-independence of data: by using random effects; and how we can control for them in a more complex manner; by using Gaussian process models. Assignment: The effect of interdependence in statistical models Part 1: Regression models are common and useful tools for understanding the relative importance of different variables on a measurement of interest. However, a key assumption of these models is that observations are independent. In the humanities, this assumption is often violated, and observations are related to each other for any number of reasons. What is the consequence of ignoring the relationships between data points? At what point should we consider using more complex methods to account for the non-independence of data? Part 2: Come up with a sample research question that could be used to exemplify the effect of non-independence. Using knowledge from your own research, or other humanities domain, develop a scenario, describe the data, and the process that causes the non-independence. Describe the scenario. Part 3: Produce an analysis with and without your independence control of choice, for your scenario described in Part 2. To produce the analyses either create your own dataset or use a pre-existing dataset. Consider simulating a dataset for more detailed control of the analysis. Write one or two paragraphs describing the results when you do not control for non-independence, and when you do control for non-independence, and one paragraph contrasting the results and reflect on how the un-controlled analysis would change your conclusion. Consider using graphs to visualise the difference. The point of this assignment is to convey your understanding of autocorrelation, when you should account for it, and the implications of not accounting for it. Cite all resources used. Some useful readings are attached. You may use them as sources within the essay. Use Chicago method of referencing uising at least 6 sources