1 Chapter 1: An Introduction to statistics

AnIntroductionToStatistics.tex or .pdf
Edward Morey - August 30, 2010
1.1 The name of this course is "Mathematical Statistics
for Economics" (Econ 7818)
So, this is a course in statistics.
What are statistics? Statistics is the plural of the word statistic.
Denition 1 A statistic is a function of one or more random variables.
To know exactly what this means, one must rst dene a variable and then
a random variable.
Put simply a variable is something that varys. That is, a variable can take
on dierent numerical values (a realized value of a variable is some number).
Each number represents a distinct state. For example, the variable G might
represent gender, where 1 corresponds to the state female and 0 corresponds to
the state not-female. Or, for example, if A is a variable such that 0 r 123 ,
A can take any numerical value between zero and 123, inclusive, the variable
A might represent age of a human. Here, I have assumed all humans are not
younger than zero and not older than 123 years (everyone would not agree on
this lower bound).
1
A is the name of the random variable (e.g. age, price, or amount of sexual
activity), and r is a numerical value of A.
2
1
From Wikipedia: The longest unambiguously documented lifespan is that of Jeanne Cal-
ment of France (18751997), who died at age 122 years and 164 days. She met Vincent van
Gogh at age 14.[1] This led to her being noticed by the media in 1985, at age 110. Subsequent
investigation found that her life was documented in the records of her native city of Arles
beyond reasonable question.[2] More evidence for the Calment case has been produced than
for any other supercentenarian case, which makes her case a standard among the oldest people
recordholders.[citation needed]. http://en.wikipedia.org/wiki/Oldest_people
2
The issue of how to distinguish between a random variable and a specic realization of
that random variable can be confusing, and the literture is not consistent in how it notationally
distinguishes between the two - I wont be either. I will try to use uppper case to denote the
name of a random variable, and lower-case to denote a specic value of that random variable.
However, I, and others, might use x to refer to both and hope the reader can determine which
is meant by the context.
1
To say that the value of a rv can be expressed with a number is not vary
restrictive: if A has cardinal properties, e.g. if A represents age, the the num-
bers have cardinal meaning; if A only has meaning in terms of a ranking, e.g.
class rank, then only the ordinal properties of the numbers are important; if A
simply denotes categories, e.g. gender or race, then the numbers mean nothing
other than dierent numbers represent dierent categories.
3
Lets assume A is a random variable (I will dene random variable in a
second). In which case
j = )(r)
, = q(r)
and
: = 4 + 7r
are each statistics, all of the same random variable, A. The letters ) and q are
the names of particular functions.
Or more generally, imagine three random variables: A, 1 and 7. In which
case
c = c(r, j, .)
/ = /(r, j, .)
,
1
= ,
1
(r, j, .)
and
,
2
= ,
2
(r, j, .)
are each statistics.
So c = :(/) is a rv.
Alternatively, one could let x refer to the rv and let x
i
refer to a value i of the rv. This
approach will be pretty clear if there is only one rv being considered. But what if there are
three rvs? Do I give them dierent names, like x, y and z? If so z
i
refers to a realized value
of z. But, what if instead of x, y and z, I had denoted the three random variables in the text
x
1
, x
2
and x
3
where the subscripts now refer to dierent rvs, not dierent observation on x.
One must be viligant.
We need to be careful and gure out what is going on by the context. Being explicit about
what we mean is also a good thing if we dont want the reader to get confused.
3
You are probably ready to conclude that I like footnotes. I do; they allow me to digressa
former student accused me of being the "King of digression"and tangents. I discuss the
properties of numbers in Morey (confuser surplus). Latent Gold, a statistical program I
use, allows one to change the specication of random variables between, cardinal, ordinal and
nominial (categorical). The dierences between many statistical and economic models are
often only the numerical properties of the dependent and independent variables.
2
All statistics are random variables, but all random variables are not statistics,
unless one denes r = r as a function.
1.2 So, what is a random variable?
Denition 2 A is a random variable if it is a variable and if it has a distribu-
tion. Said another way, A is a random variable if 8 a and / one can determine
the probability that a r / if one knows the distribution of A
Note that A takes specic values (e.g., if A is weight, each of us has a specic
weight but weight, in the population, has some distribution.)
The above denition is not self-contained. It requires that we know what a
distribution is, and we have yet to dene that term, other than we have dened
it as something that allows us to calculate Pr(a r /)
Also note the dention requires that A has a distrbution, but it does not
require that we know what distribution.
The book, Introduction to the Theory of Statistics (Mood, Graybill and
Boes) denes a continuous random variable as follows:
The variable A is a one-dimensional, continuous random variable if there
exists a function )(r) such that )(r) 0 8 r in the interval 1 r 1, and
the probability that (a r /) is
4
Pr(a r /) =
b
Z
a
)(r)dr
The function )(A) is called a density function (or a probability density func-
tion). The function, )(A), describes the distribution of A.
Any function, )(A), can serve as a density function as long as
)(r) 0, 1 r 1
and
5
+1
Z
1
)(r)dr = 1
4
Note the qualiying adjective continuous.
5
Note that f(x) 1 is not a requirement (necessary condition). It is required for certain
types of density functions, but not all of them. What types? What is required is that
b
R
a
f(x)dx 1, which follow from the restriction that
+1
R
1
f(x)dx = 1
3
Why do we care about density functions? Economic models typically assume
outcomes (how much you drink, whether the interest rate will rise) are the
result of some process with a random component: the model contains a random
variable. Or said dierently, the behavior of a variable in the model is described
by some density function.
4
Two more things:
1. A sample is the result of a random process, so a sample is a vector of
random variables. Therefore, a function of a sample is a statistic. More
on this in a bit.
2. The realized value of a random variable is not a random variable: it is a
xed number; it does not vary, so not a variable.
For example, consider the rv , age at death. Assuming the determination
of when you "kick the bucket" has a random component, is a random variable,
but once the dice is thrown and Tori "buys the ranch" a
Tori
is determined, xed,
and not a random variable. Up until that moment, a
Tori
was both a variable
and random, but neither afterwards.
6
Consider how many hours UnJung, a former student, sleeps a night; assume
its determinination has a random component. Let o denote the number of hours
he sleeps in a night, so :
t
is how many hours he sleeps on night t. Before night
t. :
t
is a rv, but once the night is over, :
t
is a xed number.
An interesting question is whether the world is inherently randomgod rolls
dice, as in quantum mechanicsor the world is deterministic and it just seems
random from our perspective because we cannot observe or measure all of the
things that determine things. While interesting, this distinction is not critical
for studying statistics in 7818.
6
"Kick the bucket" and "buy the ranch" are collequial expressions for dying. There are
hundred of collequial expressions for dying.
5
1.2.1 Make up a density function for a continuous random variable
Prove that your function is a density function.
Graph your density function.
Some student examples
6
Another example
)(r) =

0 i) r < .5
.5
round(x)
i) r .5
where ron:d(r) is dened as the integer
closest to r, with .5 rounded up.
0
0.1
0.2
0.3
0.4
0.5
0.6
1 2 3 4 5 6 7 8
x
f
(
x
)
7
Can you prove this is a density function.
Another example: choose a 0 and dene )(r) as follows:
)(r) =
8
<
:
0 r < a

3
4a
3
r
2
+
3
4a
a r a
0 r a.
The graph of this function is an upside-down parabola centered around the
j-axis, which touches the r-axis at points a and a, and crosses the j-axis at
3
4a
. Here is what it looks like if a = 1.
1 0.5 0 -0.5 -1
0.75
0.625
0.5
0.375
0.25
0.125
0
x
y
x
y
Note that it is much easy to make up a density function )(A) when A has
a limited domain, rather than an innite domain. It is tough to nd functions
where the domain is innite and the area under the function is one.
8
After one has chosen some population variable to study, how does does/should
one decide what to assume about its density function?
For example, what if the rv is number of times living individuals of have
been married.
What restrictions might one realistically impose on this distribution.
9
1.2.2 Note that we used the notion of probability to dene a random
variable
A variable is a rv if there exists some probability that the variable lies in the
interval a, /.
It is sometimes easy to forget that statstics are all about determining or
estimating probabilities.
For example, in OLS regression anlysis, something many of you know, the
probabilities are not always explicit. Many, unfortunately, xate on the OLS
parameter estimates, but the probabilities are there. For example, given the
OLS parameter estimates, what is the probability that the true value of the
parameter lies between a and /? We should be most interested in that question.
10
1.3 What do statisticans do?
Put simply, statisticians do statistics. Sam, the statistican, does statistics in
the same sense that Tiger, the golfer, does golf, but Sams does not do statistics
as well as Tiger does golf.
Theoretical statisticans and theoretical econometricans propose statistics to
better understand random processesesmost random processes are determined,
in part, by other random processess. They then try to determine the properties
of those statistics.
For example, for some random process with unknown parameters, statisti-
cans develop statistics based on samples of random variables from that process,
statistics that either help to describe the process or to our estimates of the
unknown parameters our both.
The goal is to nd a statistic that describe the process and that has desirable
properties for the question at hand. (To accomplish this goal, one has to obvi-
ously decide on what properties are, and are not, desirable - what is desirable
also depends on the question at hand.)
A statistic is like a signcant other, one wants one with desirable properties.
Applied statisticans use these desirable statistics, along with data, to esi-
mate things about the world. Implicit in the approach is the idea that observed
outcomes are the result of some random process. Remember that a statistic is a
function of random variables, and the function has parameters, which the statis-
tican wants to estimateestimation means using data to estimate the parameters
in statistics.
11
If one reads statstics books, one quickly gets the idea that statisticans have
a thing about urns and drawing balls from urns. When entering the kitchen
to make the kids breakfast, the statistican takes the lid o the "breakfast urn"
and draws a ping-pong ball. If the ball is red it is eggs, blue ceral, ..... Maybe
there is dierent urn for weekend breakfasts. Tonight will it be TV in bed or
sex with the spouse, it all depends of the draw from the bedroom urn.
1.3.1 Why doing statistics is more dicult than watching TV
A statistic is a function of random variales, so a random variable. Random
variables have density functions, so a statistic has a density function. It is often
damn dicult to deterimine that density function.
For example, imagine that A, 1 and 7 are each a random variable, such that
A has the famous Guber distribution, 1 has the famous Gomer distribution,
and 7 has the not-so-famous Snerd distribution.
Now dene the statistic
: = :(r, j, .)
= exp(tan(r
2
j) +rj(ln(.5. exp(.j)))
12
and gure out the density function for o, )(:).
1.3.2 Econometricans are, one would hope, a subspecies of statisti-
cans
Some quotes from Peter Kennedys A Guide to Econometrics
Econometrics is what econometricians do.
Econometrics is the study of the application of statistical meth-
ods to the analysis of economic phenomena.
What distinguishes an econometrician from a statistician is the
formers preoccupation with problems caused by violations of sta-
tisticians standard assumptions; owing to the nature of economic
relationships and the lack of controlled experimentation, these as-
sumptions are seldom met.
Econometricans are often criticized, and often by other econometricans.
They have a bad habit of ignoring the quality of their data. Again, some quotes
from A Guide to Econometrics
7
Econometricians are often accused of using sledgehammers to
crack open peanuts while turning a blind eye to data deciencies
and the many questionable assumptions required for the successful
application of these many techniques.
Econometric theory is like an exquisitely balanced French recipe,
spelling out precisely with how many turns to mix the sauce, how
many carats of spice to add, and for how many milliseconds to bake
the mixture at exactly 474 degrees of temperature. But when the
statistical cook turns to raw materials, he nds that hearts of cactus
fruit are unavailable, so he substitutes chunks of cantaloupe; where
the recipe calls for vermicelli he used shredded wheat; and he substi-
tutes green garment die for curry, ping-pong balls for turtles eggs,
and for Chalifougnac vintage 1883, a can of turpentine. (Valavanis)
7
Google Books at http://books.google.com/books?id=B8I5SP69e4kC&dq=Kennedy+a+guide+to+econometrics&printsec=frontcover&source=bn&hl=en&ei=t2iVSryyI4fmM7n6_PgO&sa=X&oi=book_result&ct=result&resnum=4#v=onepage&q=&f=false
13
It is the preparation skill of the econometric chef that catches the
professional eye, not the quality of the raw materials in the meal, or
the eort that went into procuring them. (Griliches
8
)
The art of the econometrician consists in nding the set of as-
sumptions which are both suciently specic and suciently real-
istic to allow him to take the best possible advantage of the data
available to him (Malinvaud)
9
The applied econometrician: The applied econometrician, unlike
the theoretical econometrician, needs to worry as much about her
data as about the theory. The forecasts and predictions generated
by the econometric model are only as good as the data that produced
them.
A well-known econometrican recently mentioned to me that he was hired
by a group of wealthy gamblers to use his choice-modeling skills to predict the
outcome of horse races. It might be important that he get it right.
10
1.4 Other perspective on statistics and statisticians
The following quote is from the front of The Advanced Theory of Statistics, Vol.
2, by M.G. Kendall and A. Stuart. They attributed it to the ctitious K.A.C.
Manderville, The Undoing of Lamia Gurdleneck.)
"You havent told me yet," said Lady Nuttal, "what it is your
anc does for a living."
"Hes an statistician." replied Lamia, with an annoying sense of
being on the defensive.
Lady Nuttal was obviously taken aback. It had not occurred to
her that statisticians entered into normal social relationships. The
species, she would have surmised, was perpetuated in some collateral
manner, like mules.
"But Aunt Sara, its a very interesting profession," said Lamia
warmly.
8
More about Zvi Griliches at Wiki: http://en.wikipedia.org/wiki/Zvi_Griliches
9
Edmond Malinvaud published in 1964, Statistical Methods in Econometrics. More details
about Edmond can be found at Wikihttp://en.wikipedia.org/wiki/Edmond_Malinvaud
10
It should not surprise that many statisticans and econometricans gamble; probability
theory developed to improve ones odds in games of chance.
14
"I dont doubt it," said her aunt, who obviously doubted it very
much. "To express anything important in mere gures is so plainly
impossible that there must be endless scope for well-paid advice on
how to do it. But don"t you think that life with an statistician
would be rather, shall we say, humdrum?"
Lamia was silent. She felt reluctant to discuss the surprising
depth of emotional possibility which she had discovered below Edgars
numerical veneer.
"Its not the gures themselves," she said nally, "its what you
do with them that matter."
Some additional quotes:
To understand Gods thoughts we must study statistics, for these
are the measure of His purpose. (Florence Nightingale)
Statistics are like a bikini. What they reveal is suggestive, but
what they conceal is vital. (Aaron Levenstein)
The rst lesson that you must learn is, when I call for statistics
about the rate of infant mortality, what I want is proof that fewer
babies died when I was Prime Minister than when anyone else was
Prime Minister. That is a political statistic. (Winston Churchill)
There are three kinds of lies: lies, damned lies, and statistics.
(Benjamin Disraeli, but sometimes attributed to Mark Twain)
To bad we cant email Florence and ask here what the hell she meant. Maybe
she mean that "casuality statistics", estimates of maimed and dead soldiers, are
a "measure of [Gods] purpose": part of Gods big plan.
15
1.4.1 How Takeshi Amemiya denes statistics
Statistics is the science of assigning a probability to an event on
the basis of experiments (Amemiya, Introduction to statistics and
econometrics, p. 2)
11
Statistics is the science of observing data and making inferences
about the characteristics of the random mechanism that has gener-
ated the data. (p. 3)
11
Takeshi Amemiya is a professor of economics and classics at Stanford. His home page is
http://economics.stanford.edu/faculty/amemiya
16
Both of statements suggest that one needs data to do statistics. I would
disagree; one can develope statistics and investigate their properties without
ever seeing any data. .
So, how does my denition dier from those of Amemiya. Mine is
more technical math-driven, Amemiyas more a what-is-it-for denition. In his
rst denition, Amemiya gives the purpose: determining or estimating proba-
bilities (consistent with my denition), and in his second denition, while not
mentioning probability, he does mention random. One might consider append-
ing my denition on the front of his denitions.
Implicit in his denitions, and in much of econometrics, is the assumption
that what we observe in the world is the result of draws from populations where
dierent outcomes have dierent probabilities of occuring. Think in terms of
drawing one or more balls from an urn, where the urn holds dierent colored
balls in dierent proportions.
Consider an urn that contains all dead smokers, pickled in brine, and one
wants to determine the probability that a smoker will get lung cancer - one
draws a sample of dead smokers from the urn.
12
One biopsies the lungs of each
to determine whether the rv Cancer, C, takes a value of 0 or 1 for that individual.
The result is a vector of random variables, c
1
, c
2
, ..., c
s
, where c
3
indicates the
cancer status of the third guy drawn. One plugs these observations, the sample,
into a statistic to estimate the probablity that a smoker will get lung cancer.
With his example in mind, consider Amemiyas denition of a random vari-
able: "A random mechanism whose outcomes are real numbers is called a ran-
dom variable." (p. 4).
13
He goes on to say, "The characteristics of a continous random variable are
captured by a density function." (p. 4 - later in the book he provides amore
technical denitions of a rv). (Note that whether one has lung cancer is not a
continuous rv, at least not given the way we dene cancer)
On to his third denition:
12
Alternatively, imagine a cemetery where all and only smokers are buried. One digs up a
bunch of the decomposing and takes from each a snip of lung tissue to see whether the smoker
had lung cancer. Here Reference the study that dug up frozen guys from WWI to see what
kind of u they had.
13
I would modify this to "whose outcomes can be expressed with real numbers." For example
one would still have a random variable if the variable was hair color and one used letters of
the alphabet, rather than numbers, to denote the dierent colors hair can take.
17
"Statistics is the science of estimating the probability distrib-
ution of a random variable on the basis of repeated observations
drawn from the same random variable."
14
(p. 4)
14
The term density function is typically only used to describe the distribution of the rv
if the rv varys continuously over some range (it is a continous rv ). If a rv can take only a
countable number of values (it is a discrete rv ), each with some probability, we dont call its
distribution a density function. Rather we call it a discrete distribution. The term probability
distribution refers to either a density function, a discrete distribution, or some combination
of the two.
18
1.5 Many statistics of interest are call estimators
An estimator is a type of statistic. Specically, it, with data, generates an
estimate of a parameter, or parameter range, in the data-generating process.
We assume members of our population-of-interest are generated by a process,
a process with a random component - a data-generating process.
And assume members of the population can be characterized in terms of
some small number of random variables. A member is simply a realization of
those random variables.
Write down a simple data-generating process:
For example, assume that the rv of interest is j and j
i
is a realization of j.
15
I am going to assume that j
i
= ar
"i
where -~(0, o)
or another data-generating process:
Assume the rv of interest is \, glasses of wine a day, and )(n) =

w
e

w!
where j 0.
If the population is all humans now alive we might be interested in how
this population varys in terms of age, gender, height and weight. We might
be willing to assume these four variables are random variables - their realized
values are generated by a random process.
Each of you can be described as realized value of that random process: you
have some age, gender, height and weight.
Ones interest in this population might be in nding someone to date. You
are writing out your application for the online dating service and have gotten
to the question about what kind of person you want to date. You want a female
between 25 and 30, over six feet tall and less than 150 pounds, but are not sure
you should be this restrictive, maybe they only occur rarely in the population
So, you ask yourself, "what is the probability someone has signed up who
is female, between 25 and 30 years old, over 6 ft. tall, and weighs less than
150 pounds." To answer this questions, you need to learn about/estimate the
joint density function for humans that have signed up for this dating service,
and then use it to determine the probability that your dream date existskeep
in mind that you might not be her dream date.
15
Note that I have broken my "rule" about uppercase for the name of the rv.
19
To say that that members of the population are generated by a
process with a random component is equivalent to saying that each
member is a draw from some density function. That density function
has some functional form and we want to estimate its form and its
parmeters.
Said another say, populations have properties and we want to come up with
estimates of those properties.
Said another way, what we observe is the outcome of a process that is driven
by parameters, and we want to estimate those population parameters.
20
1.5.1 Consider something some people do: smoke cigarettes.
One could dene a random variable c as the number of cigarettes smoked per
day by an individual, where c
i
denotes the number of cigarettes smoked by
individual i: c is a random variable and c
i
is a realized value of this random
variable.
16
We might want to learn about the distribution of this random variable in
our population of interest: determine its density function. The data generating
process is draws from that density function.
Note the term population of interest. For example, the density function
for cigarettes smoked by residents of Italy is very dierent from the density
function for cigarettes smoked by residents of the U.S. And the density function
for cigarettes smoked by foreign, male graduates students in Boulder is dierent
from the density function for all Boulder residents.
What properties, if any, must these density function have?
Can the number of cigarettes smoked take any value or must it be an integer?
Can it be a negative number? Can it be zero? Can it be 1000 a day? Someone
want to check the world record for number of cigarettes smoked in 24 hours?
Go to http://www.jimmouth.com/tv04_body.html to see some idiot smoke
159 cigarettes at once
16
Not that for many i, c
i
= 0, particularly residents of Boulder. The only people in Boulder
who smoke appear to be foreign graduate students.
21
More specically, we might want to estimate what the distribution of c would
be when cigarettes cost a dollar each, and compare this to what the distribution
would be if cigarettes cost 10 cents each: this is a problem in demand estimation.
Make sure you understand why this is a problem in demand estimation.
Econometricans want to estimate the properties of populations (humans,
smokers, interest rates, prices). We do this by taking a sample(s) from the
population - we sample random variables that describe the population.
We then propose statistics of the sampled values of those random variables ,
statistics that will hopefully be good estimators of population properties. That
is, we want to estimate population parameters. The statistics that we will
use to estimate population parameters are called estimators. We want
our estimators to do a good job of estimating the population parameter.
An estimator is a function of random variables. If one plugs particular values
of the random varialbes into the function one gets an estimate. Note the dif-
ference between estimators and estimates - estimators are functions, estimates
are realized values of an estimator - estimates are outcomes/numbers.
1.5.2 One population parameter that is often of interest is it mean
Consider some rv H. If the population is small we can sometimes observe the
whole population. If so, we can calculate (not estimate) the population mean.
But, most of the time we do not observe the whole population, so are limited
to estimating the mean of the rv H. The function that we use to estimate
the mean is an estimator. The inputs into this function are one or more rvs.
Plugging in a vector of realized values of the rvs, out comes an estimate of
mean H - remember that the mean of H is not a rv, but our estimate of it is a
rv.
Dierent realizations of the random variables will generate dierent esti-
mates of the population mean of H the estimated mean will vary from sample
to sample, have sampling variation.
For example, assume the goal is estimating the mean weight in the U.S.
population, ..
17
17
Note that we have assumed that there is a mean weight. Not all random processes have
nite means. The mean weight in the U.S. population continuously increases. Maybe it will
eventually reach innity.
22
According to WolframAlpha the mean is 180 and the median is 173. How
did they get their estimate of the mean?
We observe four weights in the population: denote the weight of the rst
person observed, n
1
, second person n
2
, third n
3
and fourth n
4
.
18
Every time we sample, we get four dierent observations: a dierent sample.
In the U.S. population there is a very large number of dierent possible samples
(dierent sets of 4 people). Let w
s
(n
s
1
, n
s
2
, n
s
3
, n
s
4
) denote sample :. w
s
is
a vector of four random variables, so any function of w
s
is a statistic.
Consider the following three statistics
e n = )(n
1
, n
2
, n
3
, n
4
) = n
1
+ 3n
2
+ (n
3
n
4
)
2
b n = q(n
1
, n
4
) =
n
1
+n
4
2
n = /(n
1
, n
2
, n
3
, n
4
) = (.25 lnn
1
+.25c
w2
)
w3
+n
4
Where did these three statistics come from? I made up three functions of the
four random variables.
I now declare each of these an estimator of .; anything can be an estimator
of anything, so what I declare is not untrue, each is an estimator of .. That
said, they may be lousy estimators of ..
Every time we plug in values from a dierent sample, we will get new esti-
mates. For example n
s
= /(n
s
1
, n
s
2
, n
s
3
, n
s
4
) = (.25 lnn
s
1
+.25c
w
s
2
)
w
s
3
+n
s
4
is the
estimate of n for sample :.
If God said that b n = q(n
1
, n
4
) =
w1+w4
2
was the best estimator of ., the
applied statistican/econometrican would always use this estimator to estimate
., no matter what sample they had collected.
19
God is either unavaible or unwilling; so, we need to decide which of all feasible
estimators is the preferred estimator (which has the most desirable properties).
To determine which is the preferred estimator, from those availible, we ask the
theorists what properties we would like an estimator to have (not all theorists
agree), and which estimator has the most of those properties.
18
Note that here the subscript refers to dierent observations on w, not to four dierent
rvs. But, that said, one could also think of them as four dierent rvs. For example in every
sample there will be a rst observation, w
1
, and this will vary from sample to sample.
19
This would be an interesting God. If she wanted to be helpful, why didnt she just tell us
!?
23
Since we can never know the true population mean of H, ., we cannot judge
an estimate of mean H by how close it is to the true value. (If we knew the true
value, we would not need to do estimation.)
We judge estimators, not estimates, this point is lost on many soulsthose
souls should be dammed to Purgatory, maybe the third level of Purgatory.
Words that come to mind when we think about the properties of an estimator
include simple, linear, unbiased (vs. biased), ecient, asymtotically unbiased,
consistent, and easy to estimate.
24
1.5.3 So, how does my description of estimators relate to good-old
ordinary least squares (OLS)?
You have to wonder.
Consider a one-parameter version of the classic linear-regression model
c
i
= c(j
i
, -
i
) = 25 +,j
i
+-
i
(1)
where c
i
is the number of cigarettes consumed by individual i, and j
i
is the price
of cigarettes for individual i (assumed a variable but not a random variable).
- is assumed a random variable (rv) and -
i
is a random draw from -. Assume
that the density function of - is normal with mean 0 and variance o
2
. , is a
parameter, not a variable, it has some xed value in the population of interest.
An estimation problem only exists because we do not know , of o
First note that - is a rv, so the c is a rv, and c(j, -) = 25 + ,j + - is a
statistic (a function of a rv).
Equation 1 describes the process by which cigarette consumption is deter-
mined. Note that a statistical model/process has been assumed: the process/model
has a random component and we have assumed the density function for this rv
belongs to the family of normal distributions.
20
We have assumed most of the estimation problem away. All that is unknown
about the population is the values of parameters , and o
2
.
21
These we want to
esimate.
We want an estimator for ,.
We will use that estimator, along with realized values of the rvs that are
the variables in the estimator, to get an estimate of ,.
In OLS, we make our estimator of , a function of a sample drawn from the
assumed population. In this case, one observation in the sample is the it/ pair
drawn, (c
i
, j
i
) - we dont observe the -
i
. A sample consists of drawn pairs:
(c
1
, j
1
), (c
2
, j
2
), ..., (c
N
, j
N
).
The OLS estimator/statistic of , is the / that minimizes
N
X
i=1
(c
i
/j
i
)
2
20
Note that this is a pretty stupid (unrealistic) model because most people smoke no ciga-
rettes, in addition no one smokes a negative number of cigarettes, so consumption cannot be
normally distributed.
21
Econometricans like to assume away most of the estimation problem. We impose a lot of
assumptions on our models, often independently of anything the data might suggest.
25
Denote this esimator /
OLS
. Every sample taken will generate a dierent /
OLS
estimate of ,. Note that /
OLS
is a rv - , is not a rv; it is a constant. Let /
s
OLS
denote the OLS estimate generated by sample :.
As applied econometricans, we often mistakinly concentrate on the obtained
esimate rather than keeping in mind that our /
OLS
, /
1
OLS
, is just one draw from
a distribution of /
OLS
.
22
That is, /
OLS
is a random variable with some density function )(/
OLS
).
Much of the work of the classical linear regression model has to do with
deriving that density function.
Once we have it, we can answer question such as "Given ,, what is the
probably that an esimate, /
OLS
, will be between (, c) and (, + c)?" Or,
of more relevance, "What is the probability that , is between (/
OLS
c) and
(/
OLS
+c)?"
So, put simply, the OLS esimator is a special type of statistic, an estimator.
And, OLS estimates are rvs with some distribution.
We like OLS esimates - when we assume the classical linear-regression model
- because we can show that the OLS estimator has nice properties: it is, if one
buys the assumptions, BLUE (a Best Linear Unbaised Estimator).
Note that what has nice properties is the estimator, not any particular
estimate generated by the estimator. Our actual OLS estimate of , often sucks.
22
b
1
OLS
is the esimate from the rst sample. I assume that most of the time we only collect
one sample. When we do simulations, we will collect many samples.
26
Note that in all of my years as an applied econometrican I have
published many models, but I have never published a paper that
report the results of a linear regression.
There is much more to econometrics than linear regressions.Consider again
urns. Urns might sound far aeld from what econometrican do, but theyre not.
Drawing a sample is akin to drawing balls from urns. Consider a sample of
c, j pairs. One could view the world as consisting of a number of urns, each
corresponding to a dierent price of cigarettes, for example, urn sixteen might
includes cigarette consumption by everyone who faces a price of $3.00 a pack
for cigarettes.
27