Week 3: Data (aka intro to EDA) (01/26/2023)

Ozlem Tuncel

otuncelgurlek1@gsu.edu

⚠️ CAUTION: DO NOT SOLELY RELY ON MY NOTES. THERE MIGHT BE TYPOS AND MISTAKES. ALWAYS TAKE YOUR NOTES!

✔️ The goal of this week is to learn how to represent data using graphs and tables.

Here are some key points:

This is a gentle introduction to EDA (exploratory data analysis) which is one of the crucial aspects of data analysis.
Modeling decisions should come last - our data determines what kind of model we are going to use.
Cow analogy: Assume that you are left in a desert. You are super hungary and suddently a cow appears from nowhere. You cannot eat the whole cow, but you can get a chunck out of it and consume. This is how theory and hypothesis are interrelated. Hypothesis is an observable implication of the theory and it is only one piece of the whole. We cannot never eat the whole cow (aka theory), but we can always take one bit (aka work one part of the theory, hypothesis).
Two main types of data: discrete or continuous.
Different levels of a variable: nominal, ordinal, interval, and ratio (has a true zero, zero means absolute lack of thing).
We have 7 different data types in R: numeric, integer, double, complex, raw, logical, and character.
Statistical significance is a dichotomy: it either exists or not.

I cannot emphasize the importance of EDA - which is something we do not see in published scholarly work. Yet, this is the fundamental aspect of your research. Often, we spent most of our time on data manipulation and exploration.
I often observe students making the following mistake: load the data and run the model. This is a no-no! Although the finished product (e.g., journal article) does not include any of the EDA steps we discussed, one should not perform any analysis to the data that they never examined carefully.
Always save excel files as .csv
Basic stat book recommendation: Wonnacoat and Wonnacoat 1990. I used this in my first year and I really like it. This was on Dr. Fix’s POLS 8805 syllabus.