Transcript Quantitative Methods/Analysis

Chapter 14
Quantitative Data Analysis
Quantitative Analysis
•
The technique by which researchers
convert data to a numerical form and
subject it to statistical analysis.
•
Quantification: The process by which data
becomes numerical
•
Example: A computer cannot read the answer
“strongly agree” so the researches assigns the
numerical value 1…and so forth.
Quantification Continued
•
Most times quantifying is important is with
ordinal-level data. Why?
•
•
Nominal-level data: The numbers you may assign
have no particular value. Women=1 and Men=0
doesn’t mean anything statistically.
Interval-level data: The numbers are already
assigned in the data. Annual income is $50,000, or
your age is 27. You’re not going to assign other
numbers.
Ordinal Quantification
•
Making sure that you’re properly
representing the data by giving it the
correct numerical status.
•
We are winning the War in Iraq.
•
•
•
•
1 = Strongly Disagree
2 = Disagree
3 = Agree
4 = Strongly Disagree
Fool-Proof…certainly not!
•
There can still be flaws. To use the previous
example:
•
We are winning the War in Iraq.
•
•
•
•
•
1 = Strongly Disagree
2 = Disagree
3 = Agree
4 = Strongly Disagree
There may be a bigger difference between 1
and 2, than between 2 and 3, which would
jeopardize our analysis of reality.
Where do we find all this?
•
When downloading a dataset, there is
often a codebook. The codebook will list
all of the questions, all of the answers,
and how each of the answers were coded
into numerical form.
Baby Steps
•
Univariate Analysis: The analysis of
ONE variable for the purpose of
description.
•
You are simply trying to “describe” that one variable
• What is the average, the frequency, the
distribution, and so on.
Distribution (or Marginals)
•
A description of the number of times the
various attributes of a variable are
observed.
•
Religious Attendance
• 57 never attend
• 67 attend once/month
• 83 attend once/week
Central Tendency
•
Mean: The average, done by summing all
values and dividing by the number of
observations.
•
Median: Represents the value of the “middle”
case in a rank-ordering.
•
Mode: Represents the most frequently
observed value.
Dispersion
•
•
The distribution of values around some
central value.
Can be very important, particularly in the
case of “outliers”.
•
Example:
• 20 observations of $20,000/year (avg. $20,000)
• 19 observations of $20,000/year AND 1
observation of $200,000/year (avg. $29,000)
Dispersion
•
Range: The distance between the lowest and
highest observed value.
•
Standard Deviation: More complex
measure in which 68% of observations are 1
standard deviation and 95% of cases are within
2 standard deviations of the average. Lower
std. deviation means the data is bunched.
Higher means it is more dispersed.
Collapsing Categories
•
•
Moving responses and observations into
more general categories.
Done occasionally to counteract high
standard deviation.
•
Can be very tricky and should always have
justification (theoretical or statistical)
•
Example: Moving “very liberal”, “liberal” and
“somewhat liberal” into the same category.
Don’t Knows
•
In all surveys you have a group of people
who reply “Don’t Know”. How do they fit
into the statistical analysis?
•
When reporting percentages and marginal's you
should report them, but it gets much more
complicated in terms of average and standard
deviation.
Bivariate Analysis
•
•
Describe a case in terms of two variables
simultaneously.
Example:
• Gender
• Attitudes toward equality for men and
women
Constructing Bivariate Tables
•
•
•
Divide cases into groups according to the
attributes of the independent variable.
Describe each subgroup in terms of
attributes of the dependent variable.
Read the table by comparing independent
variable subgroups in terms of an attribute
of the dependent variable.
Multivariate Analysis
•
Analysis of more than two variables
simultaneously.
Multivariate Analysis
•
•
Analysis of more than two variables
simultaneously.
Can be used to understand the
relationship between two variables more
fully.
Ethics and Quantitative Data
Analysis
•
•
There is always the danger of defining
and measuring variables in ways that
encourage one finding over another.
• Quantitative analysts must guard
against this.
The quantitative analyst has an obligation
to report both formal hypotheses and
informal expectations that didn’t pan out.