Transcript Quantitative Techniques

Quantitative Techniques
Lecture 1: Economic data
30 September 2004
Economic data: Outline
• How economic data are used in regulation
and competition
• Overview of methods used
• Outline of module
• Accuracy
• Good practice when you get a data set
Examples
•
•
•
•
•
Comparing costs between firms
Efficiency measurement
Calculating cost of capital
Defining a market in competition policy
Assessing the effect of merger
Methods
• Descriptive statistics:
– averages, variation, graphical views
• Measuring relationships between variables:
– Correlation and regression analysis
• Building models:
– regression, DEA, spreadsheet models
• Calibrating models: making the numbers
reflect life
This module : Overview
1) Data and its analysis
2) Random Experiment – Basic probability theory
3) Empirical and Theoretical distributions of random
variables
4) Measures of central tendency, dispersion,
skewness, etc.
5) Multivariate distributions (conditional
distribution, independence and correlation)
Overview continued
6) Sampling and Sampling distributions
7) Point and interval estimation, hypothesis
testing (comparing sample means, etc.)
8) Regression Analysis: introduction
9) Regression Analysis: violation of classical
assumptions
10) Introduction to more advanced topics.
Teaching methods
•
•
•
•
•
Lecture
Reading
Paper exercises
Group discussion
Lab exercises:
– Excel spreadsheets: basics and macros
– EViews
Coursework
• Due 14 December
• Set four weeks before
• Heavily dependent on skills developed in
labs
Types of data(1)
• Quantitative
– continuous
– discrete
• Qualitative
– shape, colour, type
Qualitative data sometimes converted to
discrete e.g. 0-1 data and vice versa
Types of data (2)
• Nominal e.g. telephone numbers, vest
number in race
• Ordinal e.g. house numbers, position in
race
• Interval e.g. Fahrenheit/Celsius
• Ratio e.g. Time to run a race,
Accuracy (1)
To assess this we need to consider data
sources:
– Company accounts
– National income accounts
– Surveys
Accuracy (2)
Were the data collected for this or another
purpose?
Do they reflect the concept accurately?
Is the dataset based on a sample of a larger
population?
Is it audited or otherwise cross-checked?
Are there any incentives for accurate
reporting?
Accuracy (3)
• What is the scope for transcription error?
• Is there any estimate of accuracy?
• Are you able to cross check?
Sources of data error
• At collection source: Clerical error,
misunderstood question, conceptual error
• The incentive to look good /bad
• Wrong units ('000s, millions, etc.), $ , £
• Sampling error
• Transcription error
• Calculation error
• Rounding
Lesson
• Assume data are error-ridden
• Use checking techniques:
– descriptive statistics, graphs
– eyeballing: do the data follow expected
pattern?
Class Exercise
• You have a set of data on the hand and feet
measurements.
• Spend five minutes looking at the data and
answering the questions
• Were the answers obvious?
• Why do people not check over the
plausibility of their data more?
Data cleaning
1. Look at suspect data:
1. absolute values, trends, relationships
2. Go back and check source when in doubt
3. Always provide your users with source of
data so they can check back
4. Correct if possible
5. Omit suspect item if it affects analysis
The dangers of data cleaning
• By eliminating data which do not conform
to your prior beliefs => bias findings in
favour of your theory
• The data no longer represent the full range
of actual experience
As long as you are honest these dangers are
usually small compared with effects of
using poor quality observations