Lecture 3 - Binary outcomes Flashcards

1
Q

For a binary outcome you can calculate:
* probability of observing an event, e.g. probability of rolling a 6.
* odds of observing an event, e.g. odds of rolling a 6.
Describe how probability and odds of rolling a 6 is calculated.

A
  • probability = p= 1/6 (there are 6 outcomes and rolling a 6 is one of them)
  • odds = p / (1-p) = (1/6) / (1-(1/6)) = 1/5
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Chi2-test
What is it used for?

A

Used for categorical data to determine whether your categorical data significantly deviates from what was expected.

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

Chi2-test
What assumptions need to be met?

A
  • Both variables are nominally measured.
  • The dependent variable is categorical.
  • ≤ 20% of the cells with the expected values can be ≤5
  • No cell has values ≤1
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Chi2-test
How is significance determined?

A

When the test statistic x^2 ≥ the critical value (found by looking up the significance level (α) and the degrees of freedom in a table).

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

Chi2-test
What do you do if the assumption that no cell has values below 1 is violated?

A

Use Fisher’s exact test

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

How to calculate degrees of freedom?

A

Df = (amount of rows - 1)* (amount of columns - 1)

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

What should you do if the amount of degrees of freedom is equal to 1?

A

A continuity correction by ‘pulling’ the actual value (werkelijke waarde W) 0.5 points closer to the expected value (verwachte waarde V):
* if W is smaller then V -> + 0.5
* if W is bigger then V -> -0.5

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Logistic regression
For what is it used?

A

Used for studying whether a dichotomous dependent variable is associated with 1 or more predictors.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Logistic regression
What assumptions need to be met?

A
  • Independency of observations. This model cannot be used to research data of repeated measurements in time.
  • No multicollinearity
  • Linearity of independent variables (i.e. predictors) and log odds
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Logistic regression
Describe the general formula of logistic regression.

A

ln(odds) = b0 + b1* X

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Logistic regression
Given the general formula ln(odds) = b0 + b1 * X, how can you calculate:
* odds to experience certain event
* probability to experience certain event
* How much do the odds increase/decrease if X increases with 5?
* 95% CI for the increase in X with 5

A
  • odds to experience certain event = exp(ln(odds))
  • probability to experience certain event = odds/(1+odds)
  • How much do the odds increase/decrease if X increases with 5? = e^(5xb0)
  • 95% CI for the increase in X with 5 = e^5(b0 +/- 1.96 x SE)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly