Lecture 1 – Data and Measurement
-Fit a model
Measurement: Relationship between the numbers and what is being measured.
Why is it important to measure?
Many ways on measuring a variable; different scales (age: adult, age group, age in
What do you want to measure? Align the questions.
Different measures will give different results. Basic issues in measurement: 1. Validity: Extent to which a measure correctly represents the concept of study
Internal validity: How well study was done
External validity: Generalize results to other situations
2. Accuracy: Measure is accurate. Close to the actual value. Do you achieve the goal?
3. Reliability: Extent to which the variable is consistent in what it is intended to measure. (E.g.:
Height of FEB student: not measure everybody, but a sample. Is it reliable? Check mean between
different samples) Measure the right thing
E.g.: Tin helmets increased injuries. Only counted injuries, not included in injured data. Measure
wrong thing: misleading statistical conclusion. Not wrong numbers but conclusion. Be clear about what you measure
E.g.: Employment went up VS unemployment went up: Both. Population increased. Talk about the
rate instead of real number. Not manipulate the situation. Organizing your data: 1. Cross-sectional: Observations at a given point in time (people, households, firms, countries)
2. Time series: Same observation different point in times (months, years) 3. Panel: Combines both. Observations over a period of time.
Hult et al. 2008 – An assessment of the measurement of performance in
international business research
Key question in IB: Why do some firms outperform others? Teach underperforming countries. Vast literature:
Performance: Important variable, DV. Do something (marketing, innovation, M&A) and
result is performance.
Inconclusive results about determinants of performance
Conclusions depend on the measurement of performance: Measure wrong things, get wrong
results. What are you going to measure?
To date, no systematic investigation as to how well IB research measures performance = gap in the
literature, contribution of the paper.
Examine: Measurement of performance Most studies do not measure performance in a manger that captures the multifaceted nature of the
construct. Measure performance: 1.Type of data source: 96 papers, some used:
Primary: collected by yourself
Secondary: collected by others
2. Type of measure: different ways to measure the ‘same’ thing
Financial performance: economic goals (E.g. profitability)
Operational performance: non-financial (E.g. innovation, productivity, satisfaction)
Overall effectiveness (Reputation)
3. Level of analysis: which level of analysis to focus on
Strategic Business Unit
Beware of: 1.Selection bias: Only selected certain firms. Difference between these firms and firms who do not
have this characteristic. Is your research question still interesting and meaningful? Check validity 2. Endogeneity: Correlation between regressor X and error term E
Lecture 2 – Data and Descriptive Statistics
Analyzing the data. Graphically and descriptive. Measurement levels: 1. Nominal: Data only tells category no ranking. E.g., Which color do you like most? Give options. 2. Ordinal: Ordered category, logical order. No differences between the values. E.g., 1st, 2nd, 3rd prize.
3. Interval: Information about differences between points on a scale. Celsius. Same gaps between
variables but doesn’t allow full calculation possibilities as with ratio. E.g., Celsius scale. 4. Ratio: Same as above but should have an absolute zero. It can be calculated. Added, deduct etc.
most room for analyzing mathematically. E.g., weight. Descriptive/Summary statistics: before actual analysis
- which players
- nature of the variable
- any problems? (E.g., min, max, negative values) Summary statistics: What are you looking at, at this stage?
1. Number of observations: Small? Can you generalize?
2. Measures of central tendency:
Mean: Average. Can be influenced by extreme observations
Median: Middle point when values are ranked in order of magnitude
Mode: Most frequent value. It can be observed. Sometimes bi/multimodal.
Nominal variable? Use the mode.
Ordinal variable? Use the median.
Interval or ratio variable? Mean or median. Depends on skewness.
Skewness? Median. (Mean does not represent the data)
Not skewed? Mean. 3. Skewness: Says something about the shape of the distribution. Deviation from normal.
Towards a positive value? It is more (longer) tailed to the left, positive.
Towards a negative value? It is more negative.
The direction to the tail is the type of skewness.
Value outside the -1 to +1 range indicate a substantially skewed distribution.