Correlation Flashcards

1
Q

What is a correlation?

A
  • Looks ar relationships or associations between variables
  • Do changes in x, relate to changes in y? (If so we can then infer)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What is an example of how correlations can relate to clinical measures and instrumentation?

A

RPE and HR Correlation
* RPE (Ask how they feel)
* HR (Machine to measure)

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

Strength of Correlations

A
  • The closer to 1 the stronger the relationship (+1 or -1)
  • Data is tight to linear line formation
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Directions of Relationships

A
  • Direct: As y goes up, x goes up (Heads Northeast)
  • Indirect: As y goes down, x goes up (Head Southeast)
  • The more horizontal the line the more likely it is 0.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What correlation tests do we do for different measures?

A
  • Interval or Ratio = Pearson Product Moment Correlation Coefficent
  • Rank or Mixed (One Rank, One Interval or Ratio) = Spearman rho rank correlation coefficent

Categorical data CANNOT be correlated. (Aka Nominal)

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

How many people do you want for a correlational study?

A

At least 30!

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

After we run the correlation coefficent test, how do we visualize the data?

A
  • Use a scatter plot!
  • Put a linear line
  • Get r^2 value
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

On the coefficent test it may have significance as a line, is this important?

A

No ignore it.

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

The weaker the relationship the harder it is to ____ interpret.

A

visually

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

Coefficent r meaures a ____ relationship only!

A

linear

Curvilinear relationship will not be described well by a correlation coefficent.

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

What is a Cubic and Quadratic Relationship?

A
  • Cubic relationship: curve to it (this leads to over and underestimating relationships); See photo attached.
  • Quadratic relationship: goes up and goes down. Prabala; Ex: Arousal graph
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

When we create a scatter plot and get a quadratic curve, what does this mean about the relationship?

A
  • There is no linear relationship
  • This does not mean that their is no relationship.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

A correlation is not a

A

proportion

  • Correlation is not a proportion (ie. .40 is not 2 x’s as strong as .20 or the difference between .50 and .60 not necessarily the same as between .80 and .90 – variable dependent)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Coefficent of determination (r^2)

A
  • % of the total variance in Y scores that can be explained by the X scores. (Tells you how much variation is explained by the other measures. Tells you the quality of measures.
  • If r=.87, r^2=.76; therefore 76% of the variance of Y is explained by X.
  • 1-r^2 is the proportion of variance not explained (24%).
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

We want to have a spread in ____ and ____ for a correlational study.

A

X and Y

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

The ____ of a correlation does not mean that the coefficient represents a strong relationship.

A

significance

17
Q

____ should not be discussed as being clinically important just because they have achieved significance.

A

Low correlations

18
Q

If we have a low correlation, what should we report?

A

Report the R^2!

19
Q

Correlation ____ Comparison

A
  • does not equal
  • Just means there is a relationship if correlation is high
20
Q

Correlation and Causation

A
  • X does not cause Y or Y does not cause X
  • Correlation: relationship
  • Cause and effect is better determines by experimental design
21
Q

Other considerations for correlations

A
  • Range of Test Values (Generalization should be limited to the range of values used to generate the correlation coefficents
  • Restriction in the Range of Scores (Need a full range of X and Y; avoid ceiling and flooring effects!)
  • Assumption of Independence (No sense correlating values together that are also use to help calculate on of the two variables compared; Ex: Gait velocity and distance, distance/time = velocity, therefore will obviously have a connection.
22
Q

What are partial correlations?

A
  • A technique that allows one to describe the relationship of more than one variable.

Example:
- How well does PT Board exam scores related to undergraduate GPA (UG GPA) and PT coursework GPA (PT GPA)?
- Some overlap of UG GPA with PT Board but also PT GPA and PT Board; in addition UG and PT GPA have some overlap

23
Q

Y =

A
  • What I am trying to understand
  • Ex: HR
24
Q

X =

A
  • What I am manipulating
  • Ex: Dosage