Empirical Research

Empirial Research Methods Lecture 1

Measurement

Level of Measurement: relationship between the numbers and waht is being measured

→ Why is it important?  Eg. Age → Adult (yes/no) ,Age groups ( 1-12, 11-20…)

Measurement Scales

Binary: Two distinct categories (Male/female, yes/no)

Nominal:         - More than two categories

• Number only tells category
• No ranking
• E.g. colours, means of transportation
• Pointless to arithmetic

Ordinal: - Ordered categories (logical order)

        -Tells nothing about differences between values

        -E.g. 1st, 2nd, and 3rd prize

Interval:         - Information about differences between points on a scale

• Equal intervals represent equal differences
• E.g. Celsius scale

Ratio: -Same, but with absolute zero

        -e.g. weight

→ What matters: you cannot perform all calculations on all variables

Basic Issues in Measurement

Validity:

• Extent to which a measure correctly represents the concept of study

• Internal: How well the study was done
• External: Generalize results to other situations

Accuracy:

• Measure close to actual value
• Getting the ‘right’ answer on average

Reliability:

• Extent to which a variable is consistent in what it is intended to measure

LOOK AT IT AGAIN, JUST HALF OF THE LECTURE

Lecture 2

The Research Process

1. Research Question

1. Observe the world

1. Theory/literature
2. Hypothesis/prediction

1. Variables

1. Collect Data

1. Measurement (Lecture 1)

1. Analyze Data

1. Graphically/ descriptively ← L2
2. Fit a model

1. And back to the beginning

Descriptive/summary statistics

Step #1

Quantitative description of main features of data

Just a summary

Before actual analysis

→ Which are the players of the game?

        Which is the nature of the variables? (Discrete, continuous)

        Do you see any problems? Things to keep in mind? (e.g. range, negative values)

→ Get a feel for your data

Summary Statistics

1. Number of observations

1. Measures of central tendency
   - Mean: arithmetic average (influenced by extreme observations
   - Median: middle point when values are ranked in order of magnitude (Relatively unaffected by extreme scores)
   - Mode: most frequent value (More than one mode, e.g. bimodal, multimodal)

→ Which measure of central tendency to use?

1. Skewness
   - Says something about the shape of the distribution
   - Deviation from normal (Normal: skew=0)
   - Symmetry of a distribution, compared to a normal
   - Kind of skew labelled according to the longer tail
   - Values outside the -1 to +1 range indicate a substantially skewed distribution

[pic 1]

1. Kurtosis
   - Says something about the shape of the distribution
   - Deviation from normal (normal/mesokurtic: Kurtosis=3)
   - Degree to which scores cluster at the tails and how pointy a distribution is (peakedness of flatness), compared to a normal
   - Leptokurtic: heavy tails and pointy (>3)
   - Platykurtic: light tails and flatter (<3)

[pic 2] [pic 3]

5. Minimum and Maximum
        -Range
                Dispersion
                Defined as: maximum value – minimum value
                Affected by extreme score

                -Interquartile Range

                        Dispersion

                        Q1: lower quartile (25%)

                        Q2: upper quartile (75%)

1. Variance and standard deviation
   Dispersion

1. Variance
2. Standard deviation

1. Square root of variance
2. How spread out the data are from the mean
3. Heterogeneity of the sample

1. ‘normal’ data (bell-shaped)

[pic 4] 

Measure of association

-Sometimes part of summary statistics
-Correlation/Correlation coefficient
        - Strength of the relationship between two variables
        -Positive or negative(sign)
        - (-1, +1)
                -Magnitude says something about the strength of the relationship (size)
        -Coefficient of +1 (-1):

                -Two variables are perfectly positively (negatively) correlated
                -As one increases, the other increases (decreases) by a proportionate amount
        -Coefficient of 0:
                -No linear relationship between two variables
                - As one variable changes, the other stays the same
        -Significance