Instructor : John Tsitsiklis .
axioms
Lecture 1
probability models and .
:
·
sample space
Probability laws
·
Axioms
-
properties follow the axiors
-
that
·
Examples
-
Discrete
-
continuous
·
Discussion
-
countable
additivity
-
mathematical subtleties .
·
Interpretations of probabilities .
# the end the lecture, about the
By of you will learn all
elements of the probabilistic model .
# probabilistic model
quantitative description situation
:
of a
,
is
a
phenomenon , or an
experiment whose outcome
uncertain . Putting together, such a model ,
involves two
key steps
:
↓ We need to describe the
possible outcomes
. This is done
of the experiment by specifying
so called sample space .
which
Then ,
we
specify probability
daw ,
assigns probability to the outcomes or
to collections outcomes .
of
# probability law tells us ; for example,
whether one outcome is much more likely than
some other outcome.
probabilities have to certain
satisfythese
basic properties
axioms
in order to be
meaningful ,
are the
of
probability theory
.
For example
;
probability cannot be
negative interestingly
,
,
axioms, but
there will be
very few they are powerful and
that lots
we can see
they
have
of consequences.
we will see that they imply many other
properties , that
were not
part of the axioms .
will couple
Then
go through overy simple examples ,
we
,
with either discrete
involving models or continuous
outcomes .
will times in this class that discrete
As
you see
many
much easier continuous models
models are
conceptually ,
involves much more
sophisticated concepts , and we will
the subtle issues that
point out
I
some
of arise.
little bit about
And
finally we will talk the
big picture .
about the role
of probability theory and it's
relation with the real world .
SAMPLE SPACE
Putting together probabilistic model a model of random
a
,
a
phenomenon or a random experiment involves two steps , ,
describe possible outcomes of the phenomenon/experiment
-
of interest
-
describe our beliefs about the likelihood of outcomes .
by specifying a
probability law .
we start
by talking just about the first step namely ,
the descripti
on of possible outcomes of the experiment .
coins
So, we
carry out an
experiment ,
we
flip a coin
,
or we
flip 5 at
the time be roll a die. Whatever, is,
same , or
may we
experiment
start by making ;
it has number
of possible outcomes , we
· list (set)
of possible outcomes ,
e
- We create a set ; e ; called sample set ; set of all
possible outcomes of our experiment
Elements of that set should have certain
properties ,
the elements should be;
① Mutually exclusive &
exhaustive
② collectively .
exclusive
mutually
if at the end of experiment I tell you that ,
this outcome
happens then it should not be possible that this
,
,
outcome also happens. At the end
of the experiment there can
only be one of the outcomes that has happened .
② Being collectively exhaustive means
something
of sets exhausts
else that's together or these elements all
the
possibilities So .
,
no matter what , at the end , you will be
able
to the took
point one
of the outcomes and say
,
that's one that
place.
To summarise :
This set should be such that at the end of the
experiment, you should be able to
always to
point
I ,
exactly + of
is the
the
possible outcomes and
say
this outcome that's occurred.
be
Physically different outcomes should ,
distinguished in the
distinct
sample space and correspond to points. But say , ,
physically different outcomes, what do we mean , we mean ,
different
relevant in irrelevant
in all
aspect perhaps
, not
different aspects .
let's look at some simple and
silly example;
it
you flip a coin , you see whether, resulted Heads or Tails .
* perfectly legitimate sample space for this experiment
-
# To which consists of two
points Heads/Tails
all possibilities
these two outcomes exhausts
Together , ,
the two outcomes are
mutually exclusive so this is very legitimate ,
,
sample space for this experiment.
suppose,
outside to check the window,
while we flip the coin,
, you
look
and then , you could say that my sample space
,
is heads and
it's raining
>
· Heads and rains
This sets consisting of four
>
-
· Heads and no rains
elements is also a
· Tails and rains
· Tails and no rain perfectly legitimate
sample space .
The
experiment of
flipping
a coin .
Elements of this sample space are
mutually
exclusive and
collectively
exhaustive and to be true and
exactly one of this outcomes is
going
will have materialized at the end of the
experiment.
so
,
which sample space is the correct one .
This sample the second one contains some irrelevant
space ,
details , so, the
preferred sample space for describing the flipping
a coin the sample space is the simpler one .
of ,
preferred
which is sort of in the right granularity given , ,
what we are
interested in
But ultimately the question of which one is the right sample space ,
?
depends on what type of question you want to answer
eg If you
: have a
theory that the weather affects the behaviour of
loins
to
then ,
in order
play with that
theory / check it out
/ so on ,
then in
want to work the second sample
you might
such a case, on
,
space
.
And this is the common
feature in all
of science ,
when ever
, you put
a
model ,
you need to decide how detailed want your model
you
to be and the level
of detail is the one that capture , those aspects that
and
are relevant
of interest to you .
-
SAMPLE SPACE EXAMPLES -
sample discrete finite examples >
: - Infinite/continuous
,
space
sets
-
Two rolls
of a tetrahedral die .
roll once
coll troice .
Sequential description .
sample space
:
·
li
2 ,3 -
X =
first roll
i
rolling die once .
discrete and finite
.
only
10possibla
continuous sample space example
:
(n , 4) such that 04 ,
· y1
yn
&
( 4)
,
·
& *
record x and
y with infinite precision .
Probability Axiom
which ones are more likely to occurr which ones areless likely to
OCCUV .
I
that my dant
⑧
what is the probability
·
will hit the centre with infinite precision ?
g I
zew
-
So , it's natural that in such
,
a continuous model ,
any individual point, should have a zew
probability.
·
for this reason , instead of
assigning probabilities to individual
points ,
we will instead
assign probabilities ,
to whole set That is.
the subsets of the
sample .
P(A) :
probability of Set A .
#//
/
/whatThe
the upper
&
half
Event : a subset of sample space .
is
probability assigned to events
-
Rules :
by convention , probabilities are always given from the
range of 0-1 .
Intuitively ,
O
probability means ,
thatwe believe that something
practically cannot happen .
that certain that an event of
and
practically
I means , we are
interest is going to happen .
These rules are called :
Axioms :
PCA) 10
nonnegativity ;
-
normalization P(-2) =
1
simple and intuitive.
-finite) additivity : Ho be strengthed later .
1) AGB)
=
4 then PLAUB) =
PLA) +P(B)