Lecture 1
Aims of this course
- To give you an appreciation of the increasingly important role of quantitative methods
in empirical research.
- To provide some useful quantitative methods that you could use in empirical
research
-test hypotheses.
-answer research questions.
-make predictions about an outcome of relevance.
- To convince you that quantitative methods are not the devil’s work.
Role of statistics in research, in this course: 5 & 6
Empirical cycle
1. Select topic
2. Focus question
3. Design study
4. Collect data
5. Analyze data
6. Interpret data
7. Inform others
Descriptive vs. inferential statistics (beschrijvende vs. inferentiële statistiek)
Two main branches of statistics:
1. Descriptive statistics
Beschrijven, organiseren, samenvatten, weergeven data.
2. Inferential statistics
Gebruik van probabilistische technieken om een sample te analyseren om ons iets te
vertellen over een populatie.
Introduction to statistics
You will be able to
- define sampling and describe the relationship between a sample and a population.
- describe the uses of descriptive vs. inferential statistics.
- describe the limits of statistics for establishing causality.
● Sample = kleine groep genomen van een populatie waarin we geïnteresseerd zijn.
● IF the sample is representative of the population, THEN we can make an "informed
guess" about population values based on sample values.
● A samples is never a perfect representation of the population, but…
- A randomly selected sample is usually representative of the population from
which it was drawn.
- Samples are usually not really random (in practice), so think about ways in
which your sample might be different from the population and give a
misleading impression about the population.
,Small word vs big world
- The proportion we calculated for the last row exactly described the proportion of men
in the “small world" of the last. There is no uncertainty about this. (Descriptive
statistics).
- This proportion was also our best guess about the proportion of men in the “big
world" of the whole room and the whole course (Inferential statistics).
- When we use small world statistics to make guesses about the big world, there is
always some uncertainty (sample is vrijwel nooit precies hetzelfde als populatie).
Descriptive vs inferential statistics
This touches upon the distinction between descriptive and inferential statistics.
● Descriptive statistics: Summarizing data from a sample. There is no uncertainty.
● Inferential statistics: Making a guess about population values, based on your
sample.
- Because a sample is never perfectly representative of the population, there is
sampling error. Thus, we always express our uncertainty when using
inferential statistics.
- Later in the course, you will hear terms like standard error and p-value, which
are both ways of expressing uncertainty when using a sample to make
guesses about the population.
Random sampling
The best way to minimize sampling error, is to have a random sample, where every
individual in the population has an equal chance to be included.
- Was our sample (of the front row) random?
→ No.
- Could there be differences between people who sit on the first row, and people who
sit in the back row, that bias our estimate of the % men in the room?
→ Interest in the course, vision.
In real research, we almost never have a random sample. We often have convenience
samples, which may be different from the population in some way (biased).
,Om een betrouwbaar sample te hebben, moet sample groot genoeg zijn. Bijv. bij onderzoek
naar non-binary mensen een groter sample, omdat de populatie vrij klein is.
What about causality? - Experimental methods
Many researchers are interested in questions of causality.
● Statistics DO NOT establish causality.
● Research designs can help establish causality.
Experimental method:
● Randomly assign people to two groups; give one group a pill, and the other group a
placebo.
● If you find a relationship between group and outcome, you can argue that this
relationship is causal because:
- You randomly assigned people to groups, so there should be no difference at
the beginning.
- You treated the groups differently, so the only difference should be the
treatment.
What about causality? - Correlational methods
Descriptive/correlational methods: “naturally occurring” or survey/questionnaire data
about things you are interested in.
You can not infer causal relationships, but you can still use the data to answer research
questions!
● You can use every statistical analysis you want to find interesting/useful
relationships!
● But keep in mind: you can never demonstrate causality.
● Spurious effects appear when two things you measured are both caused by a third
thing, that you didn't measure.
Talking about data
You will be able to
- differentiate between cases and variables.
- define the four levels of measurement.
- identify the level of measurement of several variables.
Talking about data
If you have a sample, you will want to collect some data. In social science, we often talk
about cases and variables.
● Cases can be individual people, or companies, or countries.
● Variables are properties that differ between cases.
You could each be a case, and your sex and course of study are two variables
Level of measurement
Level of measurement: what kind of values does a variable have?
For example, height can take any positive value in centimeters.
, Sex, on the other hand, can have one or two (or more) ‘values’.
These distinctions are important!
➢ Is the variable continuous (height, many different values) or categorical (groups)?
➢ If you have groups: are these groups nominal (different only by name), or ordinal
(ordered groups)?
➢ If you have continuous variables, are they at interval level (distance between
measurements is meaningful) or ratio (there is an absolute 0 point)?
Categorial vs continuous
Categorical: the variable measures whether a case belongs to one of several categories.
➔ Discrete; the variable takes exact values (whole numbers, text labels).
➔ It is not possible to fall between categories.
➔ Bijv. sex (male/female/intersex), gender
(male/female/pangender/agender/genderfluid/other non-binary), happiness on a 1-5
scale.
Continuous: Variable can take any numerical value.
➔ bijv. temperature (-273.15 to ∞(?) C), age (0 to 122 years).
NOIR level of measurement - Categorical
Some finer distinctions are made, which you can remember using the mnemonic:‘n o i r’ =
‘black’ in French.
● Nominal
- Categorical variables.
- Each category is different in name only, does not correspond to values or
order (=nominal).
- bijv. male or female or intersex.
● Ordinal
- Ordered categories.
- bijv. low to high SES; level of education; 1st place, 2nd place, 3rd place.
Noir level of measurement - Continuous
● Interval
- Continuous measurement.
- Distance is meaningful: a step from 1 to 2 is exactly as “far” as a step from 2
to 3.
- There is no true 0-point
→ Temperature in Celsius: 0 is the freezing point of water.
→ Temperature in Fahrenheit: 0 is just plain cold!
● Ratio
- Same as interval, but with a true 0-point; 0 means something!!
Aims of this course
- To give you an appreciation of the increasingly important role of quantitative methods
in empirical research.
- To provide some useful quantitative methods that you could use in empirical
research
-test hypotheses.
-answer research questions.
-make predictions about an outcome of relevance.
- To convince you that quantitative methods are not the devil’s work.
Role of statistics in research, in this course: 5 & 6
Empirical cycle
1. Select topic
2. Focus question
3. Design study
4. Collect data
5. Analyze data
6. Interpret data
7. Inform others
Descriptive vs. inferential statistics (beschrijvende vs. inferentiële statistiek)
Two main branches of statistics:
1. Descriptive statistics
Beschrijven, organiseren, samenvatten, weergeven data.
2. Inferential statistics
Gebruik van probabilistische technieken om een sample te analyseren om ons iets te
vertellen over een populatie.
Introduction to statistics
You will be able to
- define sampling and describe the relationship between a sample and a population.
- describe the uses of descriptive vs. inferential statistics.
- describe the limits of statistics for establishing causality.
● Sample = kleine groep genomen van een populatie waarin we geïnteresseerd zijn.
● IF the sample is representative of the population, THEN we can make an "informed
guess" about population values based on sample values.
● A samples is never a perfect representation of the population, but…
- A randomly selected sample is usually representative of the population from
which it was drawn.
- Samples are usually not really random (in practice), so think about ways in
which your sample might be different from the population and give a
misleading impression about the population.
,Small word vs big world
- The proportion we calculated for the last row exactly described the proportion of men
in the “small world" of the last. There is no uncertainty about this. (Descriptive
statistics).
- This proportion was also our best guess about the proportion of men in the “big
world" of the whole room and the whole course (Inferential statistics).
- When we use small world statistics to make guesses about the big world, there is
always some uncertainty (sample is vrijwel nooit precies hetzelfde als populatie).
Descriptive vs inferential statistics
This touches upon the distinction between descriptive and inferential statistics.
● Descriptive statistics: Summarizing data from a sample. There is no uncertainty.
● Inferential statistics: Making a guess about population values, based on your
sample.
- Because a sample is never perfectly representative of the population, there is
sampling error. Thus, we always express our uncertainty when using
inferential statistics.
- Later in the course, you will hear terms like standard error and p-value, which
are both ways of expressing uncertainty when using a sample to make
guesses about the population.
Random sampling
The best way to minimize sampling error, is to have a random sample, where every
individual in the population has an equal chance to be included.
- Was our sample (of the front row) random?
→ No.
- Could there be differences between people who sit on the first row, and people who
sit in the back row, that bias our estimate of the % men in the room?
→ Interest in the course, vision.
In real research, we almost never have a random sample. We often have convenience
samples, which may be different from the population in some way (biased).
,Om een betrouwbaar sample te hebben, moet sample groot genoeg zijn. Bijv. bij onderzoek
naar non-binary mensen een groter sample, omdat de populatie vrij klein is.
What about causality? - Experimental methods
Many researchers are interested in questions of causality.
● Statistics DO NOT establish causality.
● Research designs can help establish causality.
Experimental method:
● Randomly assign people to two groups; give one group a pill, and the other group a
placebo.
● If you find a relationship between group and outcome, you can argue that this
relationship is causal because:
- You randomly assigned people to groups, so there should be no difference at
the beginning.
- You treated the groups differently, so the only difference should be the
treatment.
What about causality? - Correlational methods
Descriptive/correlational methods: “naturally occurring” or survey/questionnaire data
about things you are interested in.
You can not infer causal relationships, but you can still use the data to answer research
questions!
● You can use every statistical analysis you want to find interesting/useful
relationships!
● But keep in mind: you can never demonstrate causality.
● Spurious effects appear when two things you measured are both caused by a third
thing, that you didn't measure.
Talking about data
You will be able to
- differentiate between cases and variables.
- define the four levels of measurement.
- identify the level of measurement of several variables.
Talking about data
If you have a sample, you will want to collect some data. In social science, we often talk
about cases and variables.
● Cases can be individual people, or companies, or countries.
● Variables are properties that differ between cases.
You could each be a case, and your sex and course of study are two variables
Level of measurement
Level of measurement: what kind of values does a variable have?
For example, height can take any positive value in centimeters.
, Sex, on the other hand, can have one or two (or more) ‘values’.
These distinctions are important!
➢ Is the variable continuous (height, many different values) or categorical (groups)?
➢ If you have groups: are these groups nominal (different only by name), or ordinal
(ordered groups)?
➢ If you have continuous variables, are they at interval level (distance between
measurements is meaningful) or ratio (there is an absolute 0 point)?
Categorial vs continuous
Categorical: the variable measures whether a case belongs to one of several categories.
➔ Discrete; the variable takes exact values (whole numbers, text labels).
➔ It is not possible to fall between categories.
➔ Bijv. sex (male/female/intersex), gender
(male/female/pangender/agender/genderfluid/other non-binary), happiness on a 1-5
scale.
Continuous: Variable can take any numerical value.
➔ bijv. temperature (-273.15 to ∞(?) C), age (0 to 122 years).
NOIR level of measurement - Categorical
Some finer distinctions are made, which you can remember using the mnemonic:‘n o i r’ =
‘black’ in French.
● Nominal
- Categorical variables.
- Each category is different in name only, does not correspond to values or
order (=nominal).
- bijv. male or female or intersex.
● Ordinal
- Ordered categories.
- bijv. low to high SES; level of education; 1st place, 2nd place, 3rd place.
Noir level of measurement - Continuous
● Interval
- Continuous measurement.
- Distance is meaningful: a step from 1 to 2 is exactly as “far” as a step from 2
to 3.
- There is no true 0-point
→ Temperature in Celsius: 0 is the freezing point of water.
→ Temperature in Fahrenheit: 0 is just plain cold!
● Ratio
- Same as interval, but with a true 0-point; 0 means something!!
