Introductory Statistics for Economists
ECON 2500 – Winter 2011 – Xianghong Li
Chapter 3 – Producing Data – Feb 1
- Statistics is concerned with interpreting already available data as well as
producing new data.
- This chapter will help you understand:
o Very basic principles of getting trustworthy data.
o Develop a critical view of whether existing data use trustworthy or flawed.
- Limitation of anecdotal evidence.
- Available data: from the library and the internet.
- Producing data
Introduction: Observational Study vs. Experiment
- Observational study: observes individuals and measures variables of interest.
Subjects (usually people) are not asked to do anything other than to respond
truthfully to questions.
- Experiment: deliberately imposes some treatment on individuals in order to
observe their responses.
Observational Study or Sampling
- Used more frequently in the social sciences.
- Examples include opinion polls, labour market activity surveys.
- Usually, not everyone is in a sample.
- A census includes each individual in the population. Such large samples are very
expensive. Canada conducts a census every five years.
- Used more frequently in sciences and social experiment is less common.
- Designating and actually conducting the experiment is important.
- Experiment is not foolproof. Response can be confounded (contaminated) by
lurking (overlooked) variables, and an well-intended experiment could provide
Design of Experiment – Concepts
- Experimental units or subjects.
- Treatment: experimental condition applied to a unit.
- Response variables (outcomes).
- Explanatory variables (factors).
- Level of factors: each factor can take several values in an experiment and each
value is a level for that factor.
- Each treatment is formed by combining a specific value of each of the factors. Comparative Experiments: Importance of Having a Control Group
- A design of a study is ‘biased’ if it systematically favours certain outcomes.
- Control group: the group without treatment serving for comparison. This group
will be put in the same environment (experiencing the same influence from
How to Make the Treatment and Control Groups Comparable
- Holding everything else constant, let the treatment be the only difference
between the two groups.
o Limitations: can only match on observed factor, e.g. education but not
- Randomization: rely on chance.
o Especially useful to even out the influence from unobserved
o Limitation: may not be feasible.
Principles of Experimental Design
Cautions about Experimentation
- Double blind
o Neither the subjects nor the people working with the subjects shall know
which treatment any subject had received (avoid the knowledge of the
treatment influencing the outcomes).
- Strong influence of background variables.
- Lack of realism.
Refined Experimental Design – Blocking
- Matched pairs (a special case, two subjects per block).
- Block designs in general
o Choosing similar units to form blocks.
o Random assignment is carried out separately within each block.
- Effects of