Professional Documents
Culture Documents
12
outcomes possible of number total
E in outcomes of number
) ( = E P
1 ) ( 0 E P
Probability review...
13
Probability review
Often easier to work with probability of the event
complementary to event E, not E, or E
c
P(E
c
) = 1 P(E)
Conditional probability that E occurs, given that F has
is defined by P(EF) = P(E and F) / P(F)
Similarly, P(F |E) = P(E and F) / P(E)
Rearranging yields the multiplication rule
P(E and F) = P(F)P(E |F)
If P(E|F) = P(E), then E & F are said to be independent
i.e., conditioning has no effect
Equivalently, P(F|E) = P(E), or P(E and F) = P(E)P(F)
14
15
SIA: Private health insurance
Abbott defends private health rebate J uly 17, 2006 - 3:16PM
http://www.smh.com.au/articles/2006/07/17/1152988457337.html
Health Minister Tony Abbott has hit out at critics of the Federal Government's private health insurance rebate.
"I think that there has been a consistent line of ideological hostility to the private health insurance rebate from a certain
style of academic and indeed from the Labor Party," Mr Abbott told ABC Radio today. Mr Abbott's comments came
after University of NSW professor Denzil Fiebig released a study of the rebate's effectiveness.
Professor Fiebig claimed the rebate was unsustainable and was no guarantee that people with private health
insurance would use the private health system over the public one.
But Mr Abbott said the rebate took pressure off public hospitals.
"If we didn't have the rebate we'd have a lot more pressure on the public hospital system," Mr Abbott said.
He said Professor Fiebig was part of an ideological group who would never support private health care. Giving people
choice between private and public health care was vital, Mr Abbott said.
16
SIA: Private health
insurance
17
SIA: Private health
insurance
What is the relationship between having private
health insurance (PHI) & being admitted to hospital
as a public or private patient?
Data from ABS 2001 National Health Survey was
used to derive following table of relative frequencies
Admission to hospital in last 12 months
Not admitted Admitted as
public patient
Admitted as
private patient
Totals
Have PHI 40.3 1.5 5.4 47.2
No PHI 45.3 6.6 0.9 52.8
Totals 85.6 8.1 6.3 100.0
18
SIA: Private health
insurance
Treat relative frequencies
as probabilities
Find probability of the
following events:
Have PHI
Have PHI & admitted to
hospital
Admitted as a private
patient given have PHI
Admitted as a public patient
given have PHI
Are admission as private
& have PHI independent
events?
Marginal probability
P(PHI) =0.472
Joint probability
P(PHI and admitted)
=0.015 +0.054
=0.069
Conditional probabilities
P(admitted as private| PHI )
P(admitted as public| PHI )
=0.054/0.472
=0.114
=0.015/0.472
=0.032
Independence
Admission & PHI status are related (dependent)
P(admitted as private| PHI ) P(admitted as private)
19
Auditing example
Lets draw on both topics of probability &
sampling
Background
Local auditing firm services 100 companies
10 of these companies are known to have overdue
accounts
If 2 separate firms are chosen at random
What is probability distribution for number of accounts
overdue?
20
Auditing example
009 . 0
182 . 0 809 . 0 1 overdue) two (
182 . 0
99
90
100
10
99
10
100
90
) and ( ) and ( overdue) one (
809 . 0
99
89
100
90
) | ( ) (
) and ( overdue) none (
account overdue having firm ith denote Let
2 1 2 1
1 2 1
2 1
=
=
=
=
+ =
=
=
=
=
P
e e P e e P P
e e P e P
e e P P
e
C C
C C C
C C
i
21
Sampling
Auditing example illustrates distinction between
sampling with & without replacement
Because problem specified separate firms sampling was
done without replacement
Such sampling induces dependence across events
P(e
2
) depends on type of firm sampled in first draw
Independence & hence random sampling requires
sampling with replacement
Distinction not of practical importance if population is large
Redo auditing example assuming sampling with
replacement
22
Sampling
Events in experiments such as auditing example
may be represented by probability trees
These are diagrams (that resemble trees)
Keller Ex 6.5 looks at selecting 2 students (without
replacement) from 10 students (3 F, 7 M)
23
Auditing example revisted
Assumed probability of an overdue account was
known
Enabled determination of probability distribution of
number of overdue accounts
Consider a different problem
Suppose probability of an overdue account is not known
Given sample of companies auditing firm can determine
sample proportion of these that have overdue accounts
How can this statistic be used to infer something about the
true population proportion of companies with overdue
accounts?
This is an inference problem
24
Progress report #2
In SIA admission distributions
Assign probabilities to qualitative characteristics
Public or private patient
In auditing example
Probabilities assigned to quantitative characteristics
Probabilities of number of overdue accounts
Topic of random variables
Need to introduce theoretical distributions that are
useful in representing/modelling actual data
Initially discrete distribution
Binomial