Data Analysis - informatics

Data Analysis - semester 1

Cross-sectional data

Data collected a single point in time from a sample.

Provides a snapshot of a social phenomenon.

Name an example of a cross-sectional data

World Values Survey

Representative sample

A smaller group chosen from a larger group population that reflects key characteristics

World Values Survey

a repeated cross-sectional survey
examines the evolution of social values

each country is represented by a representative sample of 1200 individuals

Observations

Numeric values, categorical codes, or scale scores representing variables of interest for a set of cases in a cross-sectional dataset.

metadata

adds meaning to data by describing it

provides descriptors such as variable names, labels, coding schemes, and measurement labels so it can be properly interpreted

microdata

data gathered from individuals

macrodata

data about social units

(countries, counties, organisations, etc)

Variable

A characteristic that varies between cases.

Can refer to facts, attitudes or social values.

Values

Numbers that represent a response category or response

Variable Name

should not be more than 8 characters.

should not have special characters.

String variable

Just text or text with numbers and signs

Numeric variable

A variable just with numbers

Numbers vs classifications

Number - number and values are the same

Classifications - the category receives a value

Technical use vs substantial use of frequency tables

Technical use - see which values and value labels the variable has

Substantial use - see how people distribute relative to variables’ values

Elements of a frequency table

Valid cases

respondents giving a response to the substantial question

missing cases

respondents refusing to answer

frequency

number of cases for each value

percent

percent of cases for each value

Valid percent

percentage for each valid value.

When the number of missing values is large, differences between percent and valid percent are high.

Cumulative percent

each percentage is added.

useful when values are ordered.

Technical use vs substantial use of a contingency table

Technical - see if filters are in place

Substantial -See how one variable is associated with another, where one is considered dependent (outcome) and the other independent (predictor).

Contingency Table

independent variable

the variable you change to see if it has an effect on the dependent variable

Concepts

Concepts are abstract terms used to describe

characteristics of social units, such as gender roles, work ethic, religiosity, progressive values, and more.

Single-item scale

closed questions.

respondents are asked to select one or more choices appropriate to their situation to a SINGLE variable

Multi-item scales

closed questions.

asked to select one choice appropriate for their situation to at least two variables.

Recoding the variable

creating a new variable by modifying an EXISTING one (e.g collapsing categories, reversing the scale, changing values)

Computing a new variable

summarising a multi-item scale by creating a NEW single variable.

Listwise

if a case has missing values, it is ignored when computing the sum of variables or the mean of variables.

less valid cases.

more missing cases.

better than pairwise.

Pairwise

if a case has missing values, we will not ignore it.

more valid cases.

less missing cases.

better when the number of cases available for analysis is very low.

Bar Chart purpose

used to display a distribution of a categorical variable or to represent a metric variable with multiple distinct values.

Bar Chart x-axis and y-axis

x-axis - categories or groups being described

y-axis - numerical values or frequencies

What does a bar chart represent from a frequency table?

Valid Percent

Histograms vs bar charts

Histograms visualize quantitative data or numerical data, whereas bar charts display categorical variables

bar chart - COMPARING and displaying data across different categories, categorical data, bars do not touch

histogram - good for continuous data, numerical data, bars touch

Clustered bar chart purpose

to visualise a contingency table with percentages within one of the variables.

shows subcategories.

Stacked bar chart

the same as a clustered bar chart but the bars are stacked on top of each other

nominal data

categories with no order
e.g. gender, ethnicity, religion

use: frequencies, percentages, bar charts

ordinal data

logical order

e.g. educational level, social class

use: frequencies, percentages, median

scale data (interval/ratio)

numeric

e.g. age, income, hours worked

use: mean, median, standard deviation, histograms

frequency

raw count (how many)

percentages

frequencies converted out of 100

cumulative percent

adds percentages progressively (should end at 100%)

doesn’t work for nominal data

bar chart

nominal or ordinal

separate bars

compares things at one point in time

histogram

scale data

bars touch

percent used for missing data

valid percent

what does valid percent exclude?

missing values

frequencies table

how often each response occurs