Statistics 2 - Regression Flashcards

(27 cards)

1
Q

What is the basic linear regression equation?

A

y = mx + c

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

What does each variable in the basic linear regression equation stand for?

A
y = outcome variable
m = slope 
x = predictor variable
c = y intercept when x = 0
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What is the basic linear regression transformation taught in this course?

A

y = B0 + B1.x1

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

What does each element in the transformed linear regression stand for?

A
y = outcome variable
B0 = intercept 
B1 = slope
x1 = predictor variable
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

How might the slope be defined narratively?

A

The change in y, for a one unit change in x.

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

How might the y intercept be defined narratively?

A

The value of y when x = 0.

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

When x is a continuous variable, what interpretation does B1 take on?

A

B1 takes on a mean difference interpretation.

That is, for a mean difference between two subjects of x = 1 unit, the mean difference in y between the subjects is the slope.

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

What are you calculating when you calculate the confidence interval for slope (B1)?

A

Confidence interval for mean differences.

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

What are you calculating when you calculate the confidence interval for the intercept (B0)?

A

The CI for the mean of a single group.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

For what type of data is the CI for the y intercept (B0) useful and for what data is it less useful?

A

Useful for binary and categorical data.

Less useful for continuous data ie B0 may be a placeholder eg cannot have height of 0cm.

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

What is the formal title for R^2?

A

Coefficient of determination.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What does R^2 measure? (Give short and long definitions)

A

The strength of linear association between x and y.

Amount of variability of data points around the regression line explained by y-values.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

What is r and what does it measure?

A

r = correlation coefficient

Measures strength and direction of relationship between x and y.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

What must you do to a log odds result for logistic regression to get an understandable actionable figure?

A

Exponentiate the log odds result to gain an odds ratio for outcome y using predictor x.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

What is effect modification (interaction)?

A

Effect modification occurs when the relationship between two variables (x and y), depends on the level of a third variable z.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

Can you test for Effect Modification through adjustment?

A

No. Adjustment is used to test for Confounding.

18
Q

How do you test for Effect Modification?

A

By comparing separate x-y estimates for different groups or levels of z eg Male vs Female, low vs med vs high

19
Q

What’s the mnemonic for calculating Risk from logistic regression results?

A

LETS LOPP branches

20
Q

What does each letter of the mnemonic for calculating Risk from Logistic Regression results stand for?

A
LETS LOPP branches
L = log y (OR) calculation for given x-value
E = exponentiate from log to normal
T = transform OR into P (= p/1+p)
S = switch P to %

Log
Odds
Prob
Percent

21
Q

What does a Spline summarise?

A

A non-linear relationship (a dog leg relationship).

Spline is ‘split-line’

22
Q

What is a Lowess plot?

A

A “LOcally WEighted Scatter plot Smoothing” line through a scatter plot to allow trends and relationships to be visualised.

23
Q

What is a Hazard Ratio?

A

A measure reported in TTE analysis summarising the chance of an outcome in a treatment group, compared to the control group.

24
Q

What is a Propensity Score?

A

PROBABILITY of being in the EXPOSED group, given certain CONFOUNDER values

25
Q

What must the predictor be in a PROPENSITY SCORE calculation?

A

Binary eg Exposed/Unexposed

26
What can you do with a PROPENSITY SCORE?
Create quartiles of B1. ie Probability of being Exposed, given population characteristics Eg Māori ethnicity could be associated with damp home, and hospital admission. But all you’d see is hospital admission and think that ethnicity was the driver.
27
Subtraction of results of comparison results on the log scale is the same as what on the normal scale? i.e log odds (A) - log odds (B)
Division
28
Addition of variables on the log odds scale is the same as what on the normal scale?
Multiplication