Riassunto esame Research methods, Prof. Binelli Chiara, libro consigliato Political Analysis, Loveless

POLITICAL ANALYSIS

Matthew Loveless

(1-4) INTRODUCTION..............................................................................................................................................2

THE SCIENTIFIC METHODS AND STATISTICS..............................................................................................2

STATISTICS: DESCRIPTION, INFERENCE AND CONTROL.......................................................................... 2

THEORY AND HYPOTESES............................................................................................................................ 3

DATA AND VARIABLES.................................................................................................................................... 4

CONCEPTUALIZATION AND OPERATIONALIZATION................................................................................... 5

DATA AVAILABILITY..........................................................................................................................................5

RESEARCH DESIGN........................................................................................................................................ 7

(6-7) DESCRIPTIVE STATISTICS: MEASURES OF ASSOCIATION I....................................................................8

DESCRIPTION AND CONCEPTUALIZATION.................................................................................................. 8

DATA, DATASETS AND VARIABLES................................................................................................................ 9

LEVEL OF MEASUREMENT...........................................................................................................................10

MEASURES OF CENTRAL TENDENCY AND DISPERSION........................................................................ 12

GRAPHING......................................................................................................................................................15

YULE’S Q........................................................................................................................................................ 16

LAMBDA.......................................................................................................................................................... 17

GAMMA........................................................................................................................................................... 18

(8-9) DESCRIPTIVE STATISTICS: MEASURES OF ASSOCIATION II.................................................................19

MEANS COMPARISON...................................................................................................................................19

SCATTERPLOT...............................................................................................................................................20

CORRELATION............................................................................................................................................... 21

BIVARIATE REGRESSION..............................................................................................................................23

REGRESSION AS MODELLING.....................................................................................................................26

TROUBLESHOOTING: FIVE ASSUMPTION..................................................................................................28

(10-11) INFERENTIAL STATISTICS: HYPOTHESIS TESTING: Χ2......................................................................29

RANDOMIZATION...........................................................................................................................................30

PROBABILITY THEORY................................................................................................................................. 30

CLASSICAL HYPOTHESIS TESTING............................................................................................................ 32

X2: CHI SQUARED......................................................................................................................................... 32

EXAMPLE - Political Gender Parity..........................................................................................................34

(12-13) INFERENTIAL STATISTICS: TESTS OF SIGNIFICANCE (T-TEST)........................................................36

NORMAL DISTRIBUTION............................................................................................................................... 36

STANDARD NORMAL AND Z-SCORES.........................................................................................................37

EXAMPLE (Calculating the probability).................................................................................................... 39

CONFIDENCE INTERVALS............................................................................................................................ 40

BINOMIAL PROBABILITY DISTRIBUTION.....................................................................................................41

T-TEST............................................................................................................................................................ 41

EXAMPLE: T-Test Hypothesis Testing......................................................................................................44

DIFFERENCE OF MEANS T-TEST.................................................................................................................44

(14-16) (OLS) MULTIPLE REGRESSION..............................................................................................................46

ELEMENTS OF A MULTIPLE REGRESSION MODEL...................................................................................47

EXAMPLE - Regression Coefficient......................................................................................................... 48

MODELLING PROCESS................................................................................................................................. 51

(17) CATEGORICAL DEPENDENT VARIABLES I................................................................................................54

BINARY LOGISTIC REGRESSION MECHANICS.......................................................................................... 54

MODELING THE RELATIONSHIP.................................................................................................................. 55

EXAMPLE.................................................................................................................................................56

(18) CATEGORICAL DEPENDENT VARIABLES II...............................................................................................57

1

(1-4) INTRODUCTION

Chapters 1-4

THE SCIENTIFIC METHODS AND STATISTICS

The scientific Method is an objective and replicable analysis of data which results in evidence which

can be used to assess proposed explanations for a relationship.

Principle Number 1: The scientific method requires a transparent and replicable description of the

research design and analysis.

While research can be guided by previous work, the researcher decide to investigate a questions or

not (it’s up to him/her)

​ → these choices have a profound effect on the results.

​ → a replicable description allows others to critique or reproduce it

N.B. Science is sometimes defended as being objective but it’s not. Efforts are made, however, to try

to make it the least subjective (in order to eliminate or minimize the influence of our own prejudices

and biases).

N.B. Difference between Normative and Positivist approaches (check the book).

Principle Number 2: The scientific method attempts to identify, isolate, and explain the relationship

under investigation.

Scientific knowledge tries to explain how the world works through the use of theory

​ → the use of theory cannot be overstated.

Principle Number 3: The scientific method seeks to derive and make appropriate inferences from the

results of our research.

Inferences is not the same as prediction: you can make predictions in a number of ways

​ → “This student did well before, so they’ll probably do well again”.

But the scientific method is about inference: going beyond surface-level prediction to explain,

generalize, and build knowledge.

STATISTICS: DESCRIPTION, INFERENCE AND CONTROL

Pag.21 2

THEORY AND HYPOTESES

Why should we use theories and hypotheses in Political Science?

A theory systematically relates possible explanations for a phenomenon or set a phenomena

​ → it explains the reason one variable affects another

The essential role of theory in Political Science is trifold:

1)​ Identifies relationships about the phenomena in which we are interested

2)​ Organizes and summarizes the work that has been done on our subject

3)​ Suggests where to look for future testing.

Together these functions guide the design of our research and suggest important considerations in

our research.

Theories are simplified explanations of relationships: they are models of a part of the world in which

we are interested in investigating

​ → But all the theories are wrong because they are simplifications.

​ The process of simplification is shaped by the assumptions we make, the inability to contain

data, any error in our measurements, …

”All models are wrong, but some are useful”

HYPOTHETICO-DEDUCTIVE METHOD

The scientific method is a derivative of the hypothetico-deductive method.

As it turns out, theories very often originate from inductive reasoning

​ → means that “It moves from specific observations to general principle”.

This is the process of theory developing: this has the additional bonus of creating space in scientific

inquiry for innovations, inventiveness, spontaneity and flair.

On the other hand, deductive reasoning moves from general to specific: is the process of applying

theory to new situations or potential data

​ → these steps constitute the process of theory testing.

In Political Science, our common knowledge is progressed by both the testing of existing theory

and the development of new theories.

CHARACTERISTIC OF A THEORY

Theories are not super complex: they tell us why something had to occur and what to expect.

There are four essential and unique characteristics about strong theories:

1)​ A plausible explanation for the relationship in which we are interested

2)​ Fertile

3)​ Parsimonious

4)​ Falsifiable

Check from page 31 to 36. 3

HYPOTHESES

In science, we never test a theory directly. Instead, we test the implications of a theory — these are

called hypotheses.

●​ Hypotheses are specific, testable statements derived from a theory.

Their purpose is to help us determine whether a theory is strong (supported by evidence) or

weak (not supported by evidence).

To conduct scientific inquiry, we need to understand two key concepts:

1.​ Independent Variable (IV); The factor that the researcher manipulates or observes to see

if it has an effect.​

It represents the “cause” in a cause-and-effect relationship.​

2.​ Dependent Variable (DV); The factor that changes as a result of the independent variable.

It represents the “effect” we are trying to measure.​

Variables are simply the concepts or elements from a theory that we want to study. They allow us to

turn abstract ideas into measurable research questions.

EXAMPLE

Theory: Spending time in nature improves mental well-being.

Hypothesis (implication of the theory): People who spend 30 minutes in a park will report higher

happiness scores than people who spend 30 minutes in an urban environment.

Independent variable (IV): Type of environment (park vs urban)

●​ Dependent variable (DV): Reported happiness score

●​

Here, the hypothesis is a testable implication of the broader theory. By testing it, we can see

whether the theory holds in this situation.

This is the central load-bearing column for the scientific method.

DATA AND VARIABLES

Data and variables are the building blocks of scientific research. While theory may guide our choices,

what to include and exclude in our research will shape our results in important ways.

Data: a collection of a singular datum combined in the form of a dataset

​ → a merely systematically collected, codified observations or information.

Data become useful when we press them into service of our research. They provide us the material

for the building blocks of scientific research: samples and variables.

In order for the use of statistics to have any power at all, we must consider two things:

1)​ the actual techniques in statistics (the selection of the most appropriate techniques, formulas

and conclusions)

2)​ the grist for the (statistics) mill; the handling of data is moot if in fact the data are garbage

”GIGO” concept - garbage in, garbage out - like, no matter the technique, the quality of output is

highly contingent on the quality of the input. 4

For any research project, the quality of data is of utmost importance.

Data must:

1)​ be representative of what we are trying to study

2)​ be what we want to measure

1) REPRESENTATION OF WHAT WE ARE TRYING TO STUDY

If you are interested in European voters, we need data representative of European voters

​ → in order to achieve our scientific goal of inference, we need to valid sample of the

unobservable total population we are interested in.

Population: is the entire set of what you wish to draw conclusions about.

Sample: is a subset of units in the population of interest (is not laziness, is necessity since is nearly

always impossible to collect data on an entire population).

The degree of confidence we can have exporting the findings from the sample will be given to us by the

statistical techniques we’ll see later.

2) DATA LIKE WHAT WE WANT TO MEASURE

The data we intend to use must conceptually and operationally capture the phenomena that we are

interested in. Variables will need to represent what we are interested in but must also represent

abstract concepts we are interested in such as force, justice, power or inequality

​ → in this case, researchers may struggle to convert data into the variables necessary to test.

CONCEPTUALIZATION AND OPERATIONALIZATION

Let’s assume we want to tackle a difficult and timely issue in Political Science such as «the effect of

inequality on democracy»​ → lots going on here.

We won’t evaluate the actual inequality but instead we’ll evaluate the best indicators

→ we devise a concept of what we are interested in and explicitly what we intend to study.

These two ideas - defines and capture - are possibly the most important steps in the scientific method.

Conceptualization: is the process of giving clear definitions to the ideas we want to study. It means

taking the messy, complex world we experience through our senses and breaking it down into specific,

well-defined observations.

Operationalization: the process of capturing / matching our concept to usable, empirical referents.

Operationalization is limited by two thing:

1)​ the quality of conceptualization

2)​ the availability or suitability of indicator

DATA AVAILABILITY

Operationalization is the process of deciding how to measure the concepts in our research. But this

process is limited by the data we have available.

There are three strategies:

1)​ Use existing, commonly accepted measures. One common strategy is to rely on existing,

widely used measures. For instance, in research on democracy, scholars often turn to

5

datasets such as the World Governance Indicators, Freedom House scores, or the Polity IV

Project. Each of these is built on the same underlying idea of democracy, but they capture

different aspects of it.

Because of this, the measures tend to move in the same direction overall, but they are not

identical. Choosing one measure over another is not trivial

​ → «Your choice will impact what you ultimately find»

2)​ Another version of data availability is sufficient operationalization, where researchers

simply work with the data that already exists, even if it is not ideal. In this case, you are

largely at the mercy of the choices and assumptions made by the people who designed those

measures before you.

3)​ A final possibility is to develop your own measure. This might involve combining existing

data in a new way or collecting entirely new information. If you go this route, you must pay

special attention to two essential qualities of your measure:

a)​ validity, meaning that it actually captures the concept you care about

b)​ reliability, meaning that it produces consistent results under the same conditions

VALIDITY AND RELIABILITY

When we study an abstract concept (like democracy, poverty, or happiness), we cannot measure it

directly. Instead, we rely on indicators — concrete pieces of data that stand in for the concept.

But not every possible indicator is equally useful. To be scientifically meaningful, the indicators we

choose need to satisfy two main criteria:

Validity (in the Sartorian sense: “makes intuitive sense”), the indicator should truly capture the concept

we want to measure. In other words, when we look at it, we should be able to say: “Yes, this really does

represent the idea I’m interested in.”

Reliability, the indicator should produce consistent results when measured under the same conditions.

If you measure it today, tomorrow, or in different places with the same conditions, you should get

similar results. Without reliability, results become noisy and unstable

Example. Concept: Happiness

Indicator choice: Asking people “How happy are you on a scale from 1 to 10?”

●​ Validity: This makes intuitive sense because happiness is a subjective feeling, so asking

people directly is reasonable.

●​ Reliability: If people answer roughly the same way when asked multiple times under similar

conditions, it’s reliable.

This shows the trade-off: a measure can be valid but less reliable, or reliable but less valid.

The best indicators are both.

✅ Key takeaway: Indicators are the bridge between abstract concepts and measurable data. They

must