Lecture 11-supplement Probability: Chapter 7.4-5
More Probability rules
-
Converted set rules
The probabilty
of events satisfies the following
rule
:
.
This is easily deduced from the usual formula for sets by after dividing both sides by
The events
are said to be mutually exclusive (disjoint) if as sets they are disjoint and hence
.
The negation of an event
is its complement
and its probabilty is given by
since
is
.
There are corresponding rules for union of three or more events, but these are best derived as needed, rather than writen down and memorized.
-
Some sample calculations
. Now we give some examples.
The experiment is to draw a card from a
card standard deck.
So,
.
Consider the event of getting a red card higher than
.
In a red suit there are
such cards and there are two red suits,
so by multiplication principle, we get
sample points.
So the probability of this event is
.
The probability of getting one of the face cards, including aces,
is clearly
and the probability of not getting any face card is
therefore
or
.
Now suppose we try to pick three cards from the deck (without replacement) then the sample space size becomes
or
.
The probability of the
event of getting at least one face
card is now a bit more complicated to determine,
since we might have to make cases of getting
exactly one two or three face cards.
On the other hand, the complement -
i.e. getting no face cards is simply the count of
picking three cards from the
non face cards or
.
Thus, in this case, it is easier to compute the probability from the complement as
or
.
It is thus essential to
look for simpler ways
of calculating the answer.
There are situations where making Venn diagrams also comes in handy.
Some poker hands
.
Consider the problem of determining the
probability of getting a straight,
which means
cards in a sequence which
are
not from the same suit
.
First, we count the sample space as
or
.
Now we think of a sequence.
It starts from the ace or 2 or 3 or ... or 10.
Thus there are 10 types of sequences.
If we ignore the restriction about the suit,
then a sequence of a given type can be chosen in
ways (each)
and if we remove the
forbidden sequences of the chosen type,
we get
.
Thus the count of "straights" of various types is
.
Thus the desired probability is:
=
.
-
If we work on the type
three of a kind
, we have
possible triples and each triple can come in four types
(by omitting one of the four identical value cards),
thus there are
possible triples.
The remaining two cards need to be mismatched.
So we choose two of the remaining
values in
ways and each can take one of the four colors, so the total count is
.
Thus the total count is
.
Thus the net probability is
a lot better than above.
Actually, it is possible to allow the remaining two cards to be a pair
and this will augment the number even more. You should analyze it!
-
Conditional probability
. Often, we have made some
observations and as a result some of the sample points
from the sample space have been removed.
As a result, our probability of an event changes.
The new probability is called a
conditional probability
.
Formally we say that
given events
by
we denote
the probability of
given that
has ocurred
or
.
For example consider that in a 5 card poker,
we have drawn an ace and
.
Then the probability of a straight is
, since no sequence of 5 cards can stretch that far.
On the other hand, if we see a 10 and an ace of different suits,
then we can fill these in into a straight in
ways and
there are
total ways of choosing 3 more cards.
Thus the conditional probability is:
.
Compare this number with the probability of straights above.
Why is it larger?
Should we have expected that?
What should we do differently,
if the ace and 10 are from the same suite?
Also note that we are calculating as if the starting 10 and ace are given and we are ignoring the
choices for them. If we throw these in, our probability goes up!
-
More examples of conditional probability
Quite often, in order that a certain event happens, another event has to happen first.
For instance, to get to class on time, you need to get up on time
(with enough time to get ready) and then you need to get into a
reasonable traffic.
Given that the probability of getting up on time is 80% and then the probability of catching a good traffic pattern is 70%,
you could argue that the probability of getting to class on time
is calculated as the product
or 56%.
What happens if you are a little late in getting up,
say the probability of this is the remaining 20%.
Perhaps, then the probability of getting a good traffic pattern
is only 40%.
Now the probability of getting to class on time is
(
)(
)=(
) or only
%.
Since, to get to class, you need to get up and fight with the traffic,
you can now calculate the combined probability of getting to class
on time as
+
=
or 64%.
-
Let us formalize these calculation as follows.
Define events E and L as getting up early and getting up late. Define events G and B as getting a good and bad traffic pattern. Thus, to get to class on time, you must have the event
or
, so the probability desired is that of the event
.
Note that the probability of getting a good traffic pattern P(G)
is not given to us at all, mainly because it changes based on
the timing.
The probability of G given E i.e. P(G|E) is 80% while
probability of G given L i.e. P(G|L) is only 40%.
The probability
is then calculated
as
.
Similarly for the other
.
The two events
and
are mutually exclusive,
so their probabilities are added!
You can imagine a further possibility where you can get up
so late that the traffic pattern does not matter!
Think what could happen to the probability then.
You should also calculate the probability of missing the class due to bad traffic!
Scratch