Vous êtes sur la page 1sur 8

1.

3: Alternative Measure of Strength of Evidence


Example: Heart Transplant Operations
15% of patients who received heart transplant operations
nationally have died. After 8 out of 10 people who
received heart transplants died at St. Georges hospital,
researchers want to know if the mortality rate is higher at
St. Georges.
Observational Units: patients
Variable: mortality rate (whether the patient died
or not)
= 0.15
= 0.8
Parameter of Interest: long run proportion of
patients dying from a heart transplant operation at
St. Georges
Null hypothesis: The long run proportion of
patients dying from a heart transplant operation at
St. Georges is 15%.
Alternate hypothesis: The long run proportion of
patients dying from a heart transplant operation at
St. Georges is more than 15%.
H0 : = 0.15
HA : > 0.15
0
-value=
1000

Interpretation: Reject the null hypothesis (because -


value is less than 0.05)
Conclusion: The long run proportion of patients dying
from a heart transplant operation at St. Georges is more
than 15%.

Note: if you reject the null hypothesis, you can just copy
the alternate hypothesis (written in words) as your
conclusion
Second Example: Further data showed that 71 out of 361
patients died at St. Georges
71
= = 0.197
361

-value= 0.007
Interpretation: Reject the null hypothesis (because p-value is
less than 0.05)
Conclusion: The long run proportion of patients dying
from a heart transplant operation at St. Georges is
more than 15%.
Effects of Sample Size
More data = more information about the population
o Know more about the center (mean) and the
spread (standard deviation)
o Reduces error when estimating values for the
population
o Narrows down range of plausible values for the
population
Alternative Measure: Standardized Statistic ( score)
Standardized statistic: How many standard
deviations the observed statistic is away from the
mean
Standard error = accounts for the standard
deviation and the sample size
Standardized Statistic:
Statistic Mean
=
Standard Error


=
Standard Error

Heart Transplant Example


= 10 = 361
= 0.8 = 0.197
= 0.15 = 0.15
Standard Error = 0.111 Standard Error = 0.019
0.80.15 0.1970.15
= = 5.86 = = 2.47
0.111 0.019
Interpreting z-values
1st example: The statistic value of 0.8 is 5.86 standard
deviations above the population value of 0.15.
For z-values bigger than 1.645 or less than -1.645 =>
strong evidence against the null hypothesis
(corresponds to a p-value = 0.05)
o Further away is from 0, the stronger the
evidence
Question: How do you get a z-value of 0?
Larger z-score => small p-value => Reject
Smaller z-score => larger p-value => Fail to reject
Bob and Tim Example
Do less people think Tim is on the left? In a survey of 30
students, 13 said Tim on the left.

1. Write the parameter of interest in words


Long run proportion of people who think Tim is on
the left

2. Write the hypotheses (in symbols and/or words)


Null hypothesis: The long run proportion of people who
think Tim is on the left is 50%.
Alternate hypothesis: The long run proportion of people
who think Tim is on the left is less than 50%.
H0 : = 0.5
HA : < 0.5
3. Set up and/or calculate the z-score (use the applet
to get the standard error)
13
= = 0.43
30

0.43 0.5
= = = 0.769
standard error 0.091

4. Use the applet to get the p-value


= 0.299
5. Give an interpretation and conclusion based on
the p-value and/or z-score
Fail to reject the null hypothesis (since the p-value is
greater than 0.05 / the z-score is between -1.645 and
1.645) and conclude
(1) the long run proportion of people who think Tim is
on the left is plausibly 50%.
OR
(2) there is insufficient evidence to say the long run
proportion of people who think Tim is on the left is
less than 50%.
Note: The standard error and p-value come from the
simulation in the plot below

Vous aimerez peut-être aussi