Transcript qualitative data analysis - the political economy of war

QUANTITATIVE DATA
ANALYSIS
Researchers convert data to
numerical form and subject them to
statistical analysis
From ch 14. Earl Babbie,
The Practice of Social Research and Lawrence
Newman, Basics of Social Research, ch. 10.
 Quantitative analysis is handled by computer
programs like SPSS.
 The task of quantification is to reduce
idiosyncratic items of information to a more
limited sets of attributes composing a
variable.
 Best to keep quantification coding based on
greater detail. Code categories can be
collapsed later anytime if lesser detail is
needed.
Begin with a well developed coding
scheme,
Develop categories to reflect dimensions
of the variable, i.e. various attributes
composing a variable
Your coding choice should match your
research purpose and be
intersubjective.
A codebook is a document that describes the
location of variable and lists the
assignments of codes to the attributes
composing those variables.
Data need to be converted into machine
readable format, every category should get
a number.
I. Univariate Analysis
 Frequency Distributions
 The number of times that various
attributes of a variable are observed in a
sample (also called marginals)
 Central Tendency
 Presenting data in the form of an
average of a measure of central
tendency.
 Central Tendency Measures
 Mean: dividing the sum of the values by the
total number of cases.
 Mode: most frequently occurring attribute.
 Median: the middle attribute in a RANKED
distribution of observed attributes, the 50th
percentile.
 These measures give the CENTER of a distribution.
Normal and Skewed
Distributions
Measures of
Dispersion/Variation
 Measures the dispersion or variability around
the center
 Remember the standard error of a sampling
distribution, measures the dispersion around the
population parameter.
 Three Measures of Dispersion
 1. Range: The distance between the highest and lowest
Value.
 2. Percentile: Tells you the score at a specific spot
within the distribution- median is 50th percentile, 25th
percentile would be the spot that has 25% below and
75% above. 95th percentile would have how many
above and how many below the spot?
 3. Standard Deviation:
 Requires a ratio level measure to calculate.
 It gives the average (mean) distance
between all scores and the mean score.
 It helps you make comparisons in
distributions- for example a low or zero
standard deviation means that all cases are
very close to or identical with the mean,
while a large standard deviation implies
that individual scores are in general quite
different to the mean
Steps in the calculation of
standard deviation
 First calculate the mean of the
distribution
 Then subtract each score from the mean
 Square the result of each subtracted
score from the mean, add them all
together to get one number
 divide that number by the number of
cases to get the “Variance”
 Square root of the variance is the
standard deviation
Calculation Standard Deviation
 Z scores are standardized scores for individual
cases calculated using the mean of the
distribution and its standard deviation
 Z= Score- MEAN / Standard deviation
 95% of the cases lie within + or – 2
standard deviations from the mean, 68% lie
within one standard deviation (+ or -) of the
mean.
II.Bivariate Relationships
 Univariate stats describe one variable in isolation,
bivariate relationships are about two variables and
whether they co-vary or are independent.
 Covariation: means things go together or that
they are associated. E.g. People who have high
values on the income variable also have high
values on the life expectancy variable.
 Independence: means there is no relationship
between variables E.g. if people with many
brothers and sisters have the same life
expectancy as those with fewer then “number of
sibilings” is independent of or not related to “life
expectancy.”
 Three techniques initially allow you to
see whether a relationship exists
between two variables
 1. A scatterplot of the two variables, to
see if the “scatter” forms a pattern
 2. A cross-tabulation or percentaged
contingency-table.
 3. A zero-order correlation or statistical
measure of association
Scatterplots
Attitude
Age Group
Under 30
30-45
46-60
61 and
older
Total
Agree
54%
27%
11%
8%
100
No Opinion
12%
40%
40%
8%
100
Disagree
8%
13%
54%
26%
100
Measures of Association: Chi-Square tells you if the relationship between
the variables is statistically significant, while Lambda (nominal level data)
or Cramer’s V (ratio level data) tells you how strong the relationship is.
The Elaboration Paradigm
 Insert a control variable into the bivariate Table
 If the relationship after the control is introduced is, in
the partials
 1. The same as the original: we call it replication
 2. One partial replicates the original but the
other(s) does (do) not, we call it specification.
 3. The original does not show a relationship but
the partial (s) does (do)- we call it a suppressor
“control” variable.
 4. The partials are weaker and the control


i) comes before the other variables , SPURIOUS
relationship (Explanation)
ii) control comes in between the two variables
(intervening variable)- INTERPRETATION
Multiple Regression Analysis
 We will look at a table in class.