Title: PS601 Quantitative Methods
1(No Transcript)
2PS601 Quantitative Methods
- Dr. Robert D. Duval
- Course Introduction
- Presentation Notes and Slides
- Version of January 15, 2008
3Overview of Course
- Syllabus
- Texts
- Grading
- Assignments
- To be turned in a single Web page
- Built in stages Keep up.
- Instruction provided
- Software
- Excel
- Stata
- NVu
4Prerequisites
- An fundamental understanding of calculus
- An informal but intuitive understanding of the
mathematics of Probability - A sense of humor
5Statistics is an innate cognitive skill
- We all possess the ability to do rudimentary
statistical analysis - in our heads
- intuitively.
- The cognitive machinery for stats is built in to
us, just like it is for calculus. - This is part of how we process information about
the world - It is not simply mysterious arcane jargon
- It is simply the mysterious arcane way you
already think
6The First Two Weeks
- Review and Setting
- The Logic of Research
- Logic
- Microcomputers
- Statistics
7Overview of Statistics
- Descriptive Statistics
- Frequency Distributions
- Probability
- Statistical Inference
- Statistical tests
- Contingency Tables
- Regression Analysis
8The Logic of Research
- A quick review of the research process
- Theory
- Hypothesis
- Observation
- Analysis
9Sample Theories
- IR - Balance of Power
- Wars erupt when there are shifts in the balance
of power - Domestic Policy
- The crime rate is affected by the economy
- Democratic Peace
- Nations with democratic regimes engage in war
less than authoritarian regimes.
10Examples of Research
- Sewage treatment plants
- Energy conservation goals
- Converting Coal fired Powerm plants to oil
- Air Quality
- Sex discrimination in raises
11(No Transcript)
12- Theory
- many are normatively driven.
-
13 Theory Hypothesis
14 Theory Hypothesis Observatio
n
15 Theory Analysis Hypothesis
Observation
16 Theory Analysis Hypothesis
Observation
17 Theory Deduction Analysis Hypoth
esis Induction Operationalization
Observation
Confirmation/ rejection
18Statistics
A Philosophical Overview
- Methods as Theory
- Doesnt see statistics as a tool
- It is the embodiment of the ideas we express
- Methods as Language
- We articulate an implicit structure when we
ascribe causation or systematic patterns. - This structure may be
- logical hence mathematical
- Relational equality/inequality
- Algebraic or geometric
19Principle organizing concepts
- The Nature of the Problem
- What are we asking?
- Measurement
- How do we have awareness of the phenomenon
- Standards for comparison
- How can we infer if what we are seeing is what we
expect to see?
20Mathematical notation
- Important mathematical notation the student needs
to know. - Summation
- For instance, the sum of all Xi from i1 to n
means beginning with the first number in your
data set, add together all n numbers. - The ? is a symbolic representation of the process
of adding up a specified series or collection of
numbers
21Mathematical notation (cont.)
- Square Roots and Exponents
- e - the base of natural logarithms
- Exponential and Logarithmic Equations
22The Base of Natural Logarithms
- Where does e come from?
- e is the base of natural logarithms
- Invented by John Napier in 1618 as a concept,
but actually calculated/derived by Jacob
Bernoulli - It is the number such that the derivative of ax
equals 1.0 - It is derived from
23Compound Interest and e
- If 1.00 is put in the bank at 100 interest,
compounded annually, its future value is 2.00. - What about compounded semi-annually?
24Demystifying e (sort of)
- So how does this translate to real life?
- Compound interest
- Where
- PV Present Value (amount deposited)
- FV future value (amount accrued)
- i interest rate (e.g .06 for 6 interest)
- k number of periods/year
- n number of years
25Levels of Measurement
- Nominal
- Dichotomous (two values)
- Ordinal
- Interval
- Ratio many times handled like interval
- For instance Levels of Measurement
26Nominal Measurement
- Nominal variables are those which can be named,
but not quantified - Religion (Protestant, Catholic, Hebrew, Buddhist,
etc) - Race (Caucasian, African-American, Hispanic,
Asian, etc) - Linguistic Group
- Marital Status (Married, Single, Divorced)
27Ordinal Measurement
- With ordinal variables, there is a rough
quantitative sense to their measurement, but the
differences between scores are not necessarily
equal. - They are thus in order, but not fixed
28Examples of Ordinal Measures
- Rankings (1st, 2nd, 3rd, etc)
- Grades (A, B, C, D, F)
- Education (High School, College, Adv degree)
- Evaluations
- Hi, Medium, Low
- Likert Scales
- 5 pt (Strongly Agree, Agree, Neither Agree nor
Disagree, Disagree, Strongly Disagree) - 7 pt liberalism scale (Strongly Liberal, Liberal,
Weakly Liberal, Moderate, Weakly Conservative,
Conservative, Strongly Conservative)
29Interval Measurement
- Variables or measurements where the difference
between values is measured by a fixed scale. - Money
- People
- Education (in years)
- Age
- Constructed Scales
30Dichotomous Measurement
- Variables that only have two values.
- May be treated as nominal, however, sometimes an
ordinal quality may exist. - Gender - male, female
- Race - black, white
- Agreement - yes, no
- T/F - true, false
- Value - high, low
- and others less easy to name
- war, no war
- vote, no vote
31Ratio Measurement
- Ratio Variables have fixed zero points.
- The percentage is a ratio variable
- part/whole\
- Feeling Thermometers
- We treat ratio and interval variables the same
- Also need an upper bound
- Although not an absolute constraint
32Statistics
- Induction about the Observable World
- A statistic is a number that provides information
about some variable of interest. - Descriptive Statistics
- Numbers that describe some aspect of the world
- Inferential Statistics
- We use inferential statistics to take information
from a sample and make some inference about a
population.
33Descriptive Statistics
- There are two main ways we describe collections
of data. - Measures of Central Tendency
- Measures of Dispersion
- These two approaches give us the ability to
describe the distribution of the data what the
data looks like.
34Statistical Tools for Describing the World -
Distributions
- Intuitive Definition
- A bunch of numbers that measure a characteristic
for a group of cases. - May be represented by a set of numbers, a graph
or picture, or even a mathematical equation.
35Measures of Central Tendency
- Measures which provide some indication of the
typical value or the 'middle' of the distribution
36Measures of Central Tendency The Arithmetic Mean
(or Average)
- The sum of all of the numbers in a set, divided
by the number in the set - Most appropriate for symmetric distributions
- Influenced by extreme values
37Measures of Central Tendency The Median
- The middle number in the data set.
- (Sort the Data...)
- The Median is the middle value if there are an
odd number of cases. - The Median is the average of the two middle
values if there are an even number of cases. - Best measure for skewed distributions
- Not very tractable mathematically!
38Measures of Central Tendency The Mode
- The most frequently occurring value.
- Used primarily for nominal data.
- The peak value of a frequency distribution is
also referred to as the mode.
39Common terms for Measures of Central Tendency
- We use the idea of measures of central tendency a
great deal in everyday language. - Average, accordance, bread-and-butter,
commonplace, Commensurate, congruent, consistent,
conventional, customary, day-to-day, everyday,
frequent, garden variety, general, habitual,
humdrum, invariably, likeness, mean, median,
medium, mediocrity, middle, middling,
nondescript, normal, ordinary, popular,
prevailing, regular, the same, standard,
stereotypical, stock, typical, unexceptional,
uniform, usual - From The Elementary Forms of Statistical Reason
by R. P. Cuzzort and James S. Vrettos)
40Measures of Dispersion
- The Range
- Range Highest value - lowest value
- Uses only two pieces of information
- Strongly influenced by the particular
observations used. - A single outlier gives a very misleading view
- For instance
- The range in the length of term in office for a
President of the United States is 30 days to 12
years.
41Measures of Dispersion
- Percentiles the point on a distribution below
which that percent fall - The 95th percentile means that you are in the top
5 - The Inner Quartile range is between the 25th and
the 75th - hence the middle 50 of the data.
42The Deviation about the Mean
- The Deviation about the Mean
-
- Indicates how far a value is from the center.
- Note that in looking at how a distribution
spreads out, we are using the measure of the
center as our conceptual foundation.
43The average of the deviations
- So it would seem to make sense to calculate all
of the deviations and find their average. - This would seem to give us a measure of the
typical amount any given data point might vary.
44The Average Deviation
- Does the average of the deviations make sense?
45Calculating the Average Deviation
Xi
1 1-3-2
2 2-3-1
3 3-30
4 4-31
5 5-32
?15 3.0 ??
46The average absolute deviation.
- Can we find the average of the absolute value of
the deviations? - Yes, but difficult to use.
47Calculating the average absolute deviation
Xi
1 1-32
2 2-31
3 3-30
4 4-31
5 5-32
?15 3.0 ?6 ABD6/5 1.2
48Fixing these deviant measures
- In order to represent variation about the mean,
we must get rid of the minus signs in a
mathematically acceptable manner.
49The standard deviation
- Square the deviations to remove minus signs
- Take the square root to return to the original
scale
50Calculating the standard deviation
Xi ( )2
1 1-3-2 4
2 2-3-1 1
3 3-30 0
4 4-31 1
5 5-32 4
?15 3.0 ?6 ABD6/5 1.2 ?10 ?(10/5) s1.414
51The Variance
- The mean of the squared deviations has some
utility as well. - Variance is what we seek to explain!
52Calculating the Standard Deviation
- The best way to calculate the standard deviation
is to use a computer. - If one is not available, try the table method.
- StDevdemo.xls (Excel)
53Population measures
- OKI lied. The formula for the standard
deviation is not quite as I described. - It turns out that the Standard Deviation is
biased in small samples. - The estimate is a little too small in small
samples. - Thus we designate whether we are using population
or sample data.
54Population vs. Sample Means
55Population vs. Sample Standard Deviations
56Frequency Distributions
- A frequency distribution is a graph or chart that
shows the number of observations of a given
value, or class interval.
57The Frequency Histogram
- To create a frequency histogram
- Determine the class interval width.
- Determine the number of intervals desired.
- Tally number of observations in each range.
- Create bar chart from class totals.
- Note that
- The X-axis represents the class interval values
- The Y-axis represents the of cases
58Example Frequency Distribution
- Develop a frequency histogram for the following
crime rate data for the 50 states. - Use the data provided in class from the US
Statistical Abstract on (US Crime Rate p.5) - A brief aside on following computer
demonstrations in class - Follow the general conceptual process the
details will come later. - Most of the detail is provided on the screen
scan it and all menu items.
59Frequency Polygon
- Same as a frequency histogram except the
midpoints of the class intervals are used - Points are connected with a line graph
- A large number of classes will make the
distribution a smooth curve if there is a large
sample size.
60Frequency DistributionsShape
- Modality
- The number of peaks in the curve
- Skewness
- An asymmetry in a distribution where values are
shifted to one extreme or the other. - Kurtosis
- The degree of Peakedness in the curve
- Continuity
- Discrete versus continuous
61Frequency Distributions - Modality
- Unimodal
- Bimodal
- See ADA scores
- Multimodal
62Frequency Distributions - Skewness
- The Third Moment about the Mean
- Right Skew (Positive Skew)
- Left Skew (Negative Skew)
63Frequency DistributionsMeasuring Skewness
- Measuring skewness alternate formula
- Normal distribution has skewness 0.0
- (Normal ranges between 3.0)
64Frequency Distributions - Kurtosis
- The Fourth Moment about the Mean
- Platykurtic
- Leptokurtic
- Mesokurtic
65Frequency DistributionsMeasuring Kurtosis
- Alternate measure of kurtosis
- Normal distributions have kurtosis 3.0
66Continuity
- Discrete distribution
- Values can take on only discrete specific values
- e.g. role of a die x ? 1, 2, 3, 4, 5, 6
- Continuous distributions
- Takes on infinitely fine values.
- Sometimes the distinction is meaningless
- i.e.
67Frequency Distributions - Types
- The Normal
- Characterized by Mean and SD
- Developed by Abraham de Moivre to describe errors
in observations in astronomy - Also called the Gaussian distribution, since
Gauss discovered the use of 2 parameter
exponential functions for distributions - The normal curve is the most ubiquitous of this
class - Describes the distribution of an infinite sum or
mean of independent randomly generated variables - You should think of the normal curve as a
fundamental law of the universe!
68Frequency Distributions - Types
69Frequency Distributions - Types
- The Normal
- The Uniform
- The Log-normal
- The Exponential
- Statistical Distributions
- t
- ?-Square
- F
70Freuency Distributions Types (cont.)
- Hyper-geometric
- Poisson
- Binomial
- Gamma
- Weibull
- Logarithmic
- Benford
71A partial list of distributions
- Discrete
- Univariate
- Benford Bernoulli binomial Boltzmann
categorical compound Poisson discrete
phase-type degenerate Gauss-Kuzmin
geometric hypergeometric logarithmic
negative binomial parabolic fractal Poisson
Rademacher Skellam uniform Yule-Simon
zeta Zipf Zipf-Mandelbrot - Multivariate
- Ewens multinomial multivariate Polya
- Continuous
- Univariate
- Beta Beta prime Cauchy chi-square Dirac
delta function Coxian Erlang exponential
exponential power F Fermi-Dirac Fisher's
z Fisher-Tippett Gamma generalized extreme
value generalized hyperbolic generalized
inverse Gaussian Half-logistic Hotelling's
T-square hyperbolic secant hyper-exponential
hypoexponential inverse chi-square (scaled
inverse chi-square) inverse Gaussian inverse
gamma (scaled inverse gamma) Kumaraswamy
Landau Laplace Lévy Lévy skew
alpha-stable logistic log-normal
log-logistic Maxwell-Boltzmann Maxwell
speed Nakagami normal (Gaussian)
normal-gamma normal inverse Gaussian Pareto
Pearson phase-type polar raised cosine
Rayleigh relativistic Breit-Wigner Rice
RosinRammler shifted Gompertz Student's t
triangular truncated normal Tweedie type-1
Gumbel type-2 Gumbel uniform
Variance-Gamma Voigt von Mises Weibull
Wigner semicircle Wilks' lambda - Multivariate
- Dirichlet Generalized Dirichlet
inverse-Wishart Kent matrix normal
multivariate normal multivariate Student von
Mises-Fisher Wigner quasi Wishart - Miscellaneous
- bimodal Cantor conditional equilibrium
exponential family Infinite divisibility
(probability) location-scale family
marginal maximum entropy posterior prior
quasi sampling singular unimodal
72Charts and Graphs
- A picture is worth
- Charts convey a large amount of specialized
information in a compact way - They do not require the same type of cognitive
processing that words and numbers do - Learn to use them!
73Graphs Charts - Types
- Descriptive Graphs
- Bar Chart
- Pie Graph
- Line Graph
- Distributions
- Histogram
- Box Plot
- Steam and Leaf
74Bar Charts
- Best for displaying actual values.
- Can handle moderate of cases (bars)
- Excel calls it a column chart
75Bar Chart An Example
76Pie Charts
- Best used with small number of categories or
cases to display - Good for showing relative distribution
- Percentages, proportions
- Use only one column of data
- Plus one column of labels
77Pie Chart Examples
78Line graphs
- Best for showing data across time
- Always give dates
- Label X axis
- Indicate units on Y axis
- Use legend for multiple lines
79Line Graphs - Example
80Box Plot
- Quick picture of a distribution
- Parts of box give distribution characteristics
- Your Text is not quite accurate!
81Stem and Leaf Plot
- Good for showing distribution while preserving
data - Figuring out stems can be tricky
82Probability
- Before starting a discussion of the normal curve,
a couple of brief points about probability is in
order - Probability is essentially How likely is it that
some event x will occur? - All Probabilities range between 0.0 and 1.0
- Probabilities outside that range are impossible
they are not probabilities - We symbolize probability as P(x)
- The Probability of some event x is P
- 0.0 ? P(x) ? 1.0
83Probability Density Functions
- A probability density function is a frequency
distribution whose area is set equal to 1.0. - Most distributions are PDFs.
- They let us assess the likelihood or probability
of cases taking on particular values.
84The Normal Distribution
- The normal distribution is one of the most
- Popular
- Ubiquitous
- Useful
- distributions that we have.
- It gives great predictive ability when we can
apply it to data.
85The Normal Distribution the Formula
- The normal curve is described by the following
formula.
86The Normal Distribution (cont.)
- This formula will give us the following
distribution
87Using the Normal Distribution
- The normal distribution is a tool for examining
how values are distributed. - Think of the normal distribution as an underlying
physical process that generates values according
to a general pattern. - Because these values have this pattern, we have
information about them that translates to
probabilities.
88Standard Normal Distribution
- The standard normal distribution is specialized
example of the Normal Dist. - The Standard Normal distribution has ? 0.0 and
? 1.0 - We say this symbolically as
- Z ?N(0,1)
- (or Z is normally distributed with a mean of zero
and a standard deviation of one)
89The Normal PDF
- Because the standard normal curve is a PDF, we
can use it to make probability assessments about
values in the distribution. - We can, in essence, convert our regular normal
distribution to the Standard Normal, and examine
areas under the curve to calculate probabilities.
90Using the Normal PDF
- We know the following facts
- Area under the curve 1.0
- Its symmetric, so the probability of Xi being
greater that 0.0 is .5 - Symbolically,
- P(Xi gt 0.0) .5
91Using the Normal PDF
- We can use this information in the following
fashion - The P(0.0 ? Xi ? 1.0) .3413
- Thus 68 of the Xis will fall between ?1?
- Thus 96 of the Xis will fall between ?2?
92Standard Normal Variables
- A Standard Normal Variable is one that has been
transformed by the following formula - All Z-scores, as they are called, will have a
mean 0.0 and s 1.0
93Creating indices
- Z-score transformations are useful for adding
apples and oranges - They transform similar variables to the same
scale. - As a result, we can add together things that
would otherwise be impossible to combine.
94The assumption of normality
- As a result of the application of the standard
normal distribution, we can make probability
statements about data. - (A probability statement is one that is of the
general How likely is it that? format.) - If we can assume that the data is normally
distributed, then we know how likely it is that
individual cases are above or below selected
values expressed in standard deviation units.
95The Central Limit Theorem
- The Normal Distribution pops up in one very
important context - The Central Limit Theorem
- This is a fundamental concept that allows us to
infer the characteristics of a population based
upon a sample.
96Sampling Distributions
- The probability distribution of a statistic is
called its sampling distribution. - If we collect a sample and calculate the mean,
that is one data point in the sampling
distribution of the sample mean. - If we do this many times, we have a sampling
distribution, which we can then describe.
97The Central Limit Theorem
- The CLT tells us
- As the sample size n gets larger, the sampling
distribution of the sample mean can be
approximated by a normal distribution with mean
of ?, and a standard deviation equal to - where ? and ? are the population
characteristics.
98The Implications of the Central Limit Theorem
- We can use the CLT to make probability statements
about the sample mean because we know its
distributional characteristics. - Even if the original variable X is not normally
distributed, the sampling dist of the sample mean
is!
99Statistical inference
- We can use information about the way variables
are distributed to make assessments of
probability about them. - Many of these questions are phrased as Is A
greater (or less) than B? - This may also be phrased
- Does A belong to the same population as B?
100Assessing probabilities
- Take income
- Would you expect a doctor to have a higher income
than the population at large? - Would dog-catchers be lower?
- Would you expect males to have a higher income
than females - Is WV income lower than the national average?
What about Oklahoma?
101Statistical Decision-making
- Many of these questions are best answered with a
statement of statistical confidence a
probability assessment. - This statistical confidence places a decision
within an objective framework. - If we define the criteria for making decisions
according to some reasonable standards, then we
can remove (or certainly reduce the subjectivity
of the researcher.
102Statistical D-M (cont.)
- If you collected the following information, would
you conclude that males had a higher income than
females? - MeanMales 55.5K, MeanFemales 54.9K
- MeanMales 55.5K, MeanFemales 50.9K
- MeanMales 55.5K, MeanFemales 34.9K
- Where would you draw the line?
- Does sample size matter?
103Statistical Decision-making problem setup