The Analysis of Categorical Data- The Binary Analog of the unpaired-t-test: CHI-SQUARED TEST Flashcards

1
Q

What is Chi-squared test

A

Is a statistical hypothesis test used in the analysis of CONTINGENCY TABLES

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

How to calculate Chi squared

A
  1. Find the difference between the observed (o) and expected (e) values
  2. Take the square of that number and divide it by the expected value
  3. Add all of these calculated values from various categories to get chi squared
    If the two populations proportions are equal therefore the null hypothesis is true
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What is the null hypothesis for chi squared

A

There is no difference between the proportion of the population possessing the attribute in question to another

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What is the P value of the chi squared test

A

PROPORTION OF THE 1000X2

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

How does statistical packages compute x2 and corresponding P value

A

Through ASYMPTOTIC APPROXIMATION- this generates a large number of tables used to generate the x2 statistics

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Can you do chi-squared with larger contingencies tables?

A

Yes you can do chi squared with larger contingency tables other than 2x2 with the same formula

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What are the three important things to remember when doing a chi squared test (x2)?

A
  1. MAKE SURE THE ENTRIES IN THE TABLE YOU ANALYSE ARE COUNTS, NOT PROPORTIONS (Percentages)
    - Shouldn’t apply proportions due to standard error as when comparing with another proportions the test will be much more sensitive to departures from the value for this population in the case with large size
  2. MAKE SURE THE ENTRIES IN THE TABLE YOU ANALYSE COUNT INDEPENDENT ENTITIES
    -Doesn’t work if they aren’t independant as it could skew the x2 value and the P-value
    - make sure that the total number in any table is EQUAL to the total number of independent units in the analysis
  3. TABLES WITH SMALL EXPECTED VALUES OF LESS THAN 5 DO A FISHER’S EXACT TEST
    - As the P-value from the X2 statistic may NOT BE RELIABLE
How well did you know this?
1
Not at all
2
3
4
5
Perfectly