Data Analytics for Social Science Comparing through visualisation

Size: px

Start display at page:

Download "Data Analytics for Social Science Comparing through visualisation"

Claud Peters
5 years ago
Views:

1 Data Analytics for Social Science Comparing through Johan A. Elkink School of Politics & International Relations University College Dublin 3 October 2017

2 Outline 1 2 Embellishments and aesthetics 3 factors 4

3 Graphs are typically used to show either the distribution of one variable or the correlation between two variables. They can also be used for showing the relationship between multiple variables at once. Or between raw data and model estimations.

4 source

5 Example: income distribution source

6 Example: income distribution source

7 and ggplot Three main methods to split : aesthetics: Vary aesthetics by group, e.g. color or point size. geom_bar(aes(x = party, fill = dublin)) overlays: Add multiple geometries to the same plot, e.g. scatter plot with smooth curves. geom_point(position = "jitter") + geom_smooth(color = "black") facets: Split the graph in sub to show effects for different categories. geom_smooth(color = "black") + facet_wrap(~ party) ggplot2 cheat sheet

8 Facets myplot <- ggplot(ie, aes(x=knowledge, y=turnout) + geom_point() + geom_smooth(color = "blue") Facets in a grid of columns: myplot + facet_grid(interest ~.) Facets in a grid of rows: myplot + facet_grid(. ~ interest) Facets in a rectangular grid: myplot + facet_wrap(~ interest) source R Cookbook: Facets

9 Outline 1 2 Embellishments and aesthetics 3 factors 4

10 One can visualise different groups or effects by varying the color, shape, or thickness of lines and points. pl <- ggplot(ievoters, aes(x = interestpolitics, y = supportlabour)) pl + geom_bar(color = "blue") pl + geom_points(fill = "yellow") pl + geom_points(aes(shape = factor(party))) pl + geom_points(size = knowledge) pl + geom_point(shape = 21, colour = "black", fill = "white", size = 5, stroke = 5) R Cookbook: Colors ggplot2 docs ggplot2 aesthetics

11 Shading Sometimes shading or transparency can to overlay plots, or simply to make them look more professional. ggplot(diamonds, aes(depth, fill = cut, colour = cut)) + geom_density(alpha = 0.1) + xlim(55, 70) source

12 Titles and labels Proper titles and labels are essential. A graph should be interpretable without reading the accompanying description in the text (but the text should always describe all!). pl <- ggplot(ievoters, aes(x = interestpolitics, y = supportlabour)) pl + ggtitle("pol. interest and support for Lab") + theme(plot.title = element_text(lineheight=.8, face="bold")) R Cookbook: Titles

13 Legends bp <- ggplot(ees, aes(x = factor(dublin), y = age, fill = factor(dublin))) + geom_boxplot() bp + scale_fill_discrete(name="residency", breaks=c("1", "0"), labels=c("dublin", "Outside Dublin")) R Cookbook: Legends

14 Internal labels Sometimes it helps to add labels inside the plot itself too. pl + geom_points() + geom_text(aes(label = factor(dublin)))

15 Good captions, including sources and interpretation, is crucial for clear communication with.

16 Outline 1 2 Embellishments and aesthetics 3 factors 4

If healthier patients take a drug and sicker patients do not, we can find an association between drug and recovery even when the drug does not work.

17 If healthier patients take a drug and sicker patients do not, we can find an association between drug and recovery even when the drug does not work. If sicker patients take a drug and healthier patients do not, we might not find an association between drug and recovery even when the drug works. association causation

18 Causation For something to be a causal relationship, the cause should precede the effect in time; the dependent and independent variables should correlate; no third variable should explain this correlation; there should be a logical explanation or causal mechanism linking the two.

19 Control variables To isolate the effect of the treatment on the effect we need to control for the confounder. can be used to control, evaluating the effect of the treatment on the effect for different categories of the confounding variable.

20 Control variables: example Turnout Knowledge No Yes Low High (fake data)

21 Control variables: example High interest in politics Turnout Knowledge No Yes Low High 8 92 Low interest in politics Turnout Knowledge No Yes Low High (fake data)

22 Brexit example: attention effects prointegration attention ggplot(brexit, aes(x = attention, y = prointegration)) + geom_jitter(alpha =.2) + geom_smooth(method = "lm", se = FALSE)

23 Brexit example: attention effects Conservative Labour Liberal Democrat prointegration Other UKIP Would not vote attention

24 Conservative Labour Liberal Democrat Other UKIP Would not vote Brexit example: attention effects prointegration attention ggplot(brexit, aes(x = attention, y = prointegration)) + geom_jitter(alpha =.2) + geom_smooth(method = "lm", se = FALSE) + facet_wrap(~ party)

25 Outline 1 2 Embellishments and aesthetics 3 factors 4

26 Project 1: assignment Using the data used in Labs 1 4, 1 Take either vote or prointegration as the dependent variable. 2 Select a key independent variable of interest in the data, e.g. education, pro-integration attitude, levels of uncertainty, attitudes towards immigrants or the economy. 3 Identify potential confounding factors (among the remaining variables in the data set). 4 Use one bivariate and at least two multivariate to investigate the relationship between the key independent variable and the dependent variable, taking into account the confounders. 5 Write a 1,000-1,250 word essay (excluding captions and bibliography), including the, discussing your interpretation of the analysis.

27 Project 1: essay structure All assignments should be submitted electronically to consisting of either: a PDF file containing the essay and an R- or Rmd-file with all the commands necessary for the analysis, or a PDF file generated using RMarkdown with the essay and all necessary commands and results. Each essay should consist of 1 a short introduction, 2 a description and motivation of the data and methods used (approximately 25% of the essay), 3 the analysis including necessary and tables (approximately 35%), and 4 an interpretation and conclusion (approximately 40%). Everything needs to be properly referenced.

28 Project 1: practicalities A good starting point for references is probably: Best is to make a separate Markdown file for the analysis required for the assignments, which can be copy/pasted and then edited from the lab sheets. Recommendation: the deadline is 17 October, 1 pm, but you should have everything now to do the assignment, so get this out of the way before other modules give assignments.

STAT 113: Statistics and Society Ellen Gundlach, Purdue University. (Chapters refer to Moore and Notz, Statistics: Concepts and Controversies, 8e)

STAT 113: Statistics and Society Ellen Gundlach, Purdue University (Chapters refer to Moore and Notz, Statistics: Concepts and Controversies, 8e) Learning Objectives for Exam 1: Unit 1, Part 1: Population