MEASURING THE DIFFERENCES BETWEEN PROPORTIONS CONFIDENCE INTERVALS Flashcards
Why present Confidence intervals
It measures the likely difference between proportions is probably more important
If two groups are independant or unpaired what is required
The X2 test (chi squared test)
What are the ways to measure confidence intervals?
- ABSOLUTE DIFFERENCE
D=π1-π2,
R=π1/π2
OR=(πr/(1-π1))/(π2/(1-π2)) - THE RELATIVE RISK
- THE RATIO
What is the Null hypothesis
D=0, R=1, OR=1 are. all the same
Can you work out the probability from the odds or visa versa
Yes the probability and odds of an event are EQUIVALENT ways of QUANTIFYING THE SAME DEGREE OF UNCERTAINTY: IF YOU KNOW ONE THEN YOU CAN FIND THE OTHER
For example:
If an event has probability π, then the same event has odds π(1-π) or π/(1-π)
So..
Probability of π=2/3 then odds is (2/3)/(1-2/3)=2.
Odds (n), the probability will be n/(1+n)
What has the probability have to always been between
0 and 1
What does odds have to be
Bigger than 0
What is the 95% confidence intervals found as
D±1.96 x standard error
What is the 99% confidence intervals found as
D± 2.58 x standard error
Is there a connection between any odds ratio greater than one between two groups
Yes there is a connection between any odds ratio GREATER THAN ONE and its reciprocal, which is between 0 and 1
Are all the values from 1 up to infinity equivalent to the values BETWEEN 0 AND 1 and does it give the distribution of the Odds ratio (OR) a natural skewness
Yes all values from 1 upto infinity are in some ways equivalent to the values between 0 and 1 GIVES DISTRIBUTION of the OR A NATURAL SKEWNESS
How to work out the standard error of logOR
It is the SQUARE ROOT OF THE SUM OF THE RECIPROCALS OF THE ENTITIES IN THE 2X2 TABLE UNDERLYING THE RATIO
What is the natural logarithm for OR
0.8019
How to find the confidence intervals for the OR
Confidence intervals for the odds ratio (OR) is found by TAKING THE NATURAL ANTILOG OF THE VALUES (sometimes written as the EXPOTENTIAL of the values)
Who do you find the mid-point inbetween the confidence intervals
You do 1/2 (first interval+ second interval)
This will be larger than the Odds Ratio (OR)