Research Methods: 2 M.Sc. Physiotherapy/Podiatry/Pain

1
Research Methods 2M.Sc. Physiotherapy/Podiatry/P
ain

• Correlation and Regression

2
Relationships Between Variables

• Exploring relationships between variables
• What happens to one variable as another changes

3
Relationships Between Variables

• Correlationthe strength of the linear
  relationship between two variables.
• Regression the nature of that relationship, in
  terms of a mathematical equation.
• In this module we are only concerned with linear
  relationships between variables.

4
Correlation
5
Correlation
Correlation Coefficient r -1 ? r ? 1
6
Correlation r 1, perfect positive correlation
7
Correlation r -1, perfect negative correlation
8
Correlation 0 lt r lt 1, positive correlation
9
Correlation -1 lt r lt 0, negative correlation
10
Correlation r ? 0, no linear relationship
11
Correlation r ? 0, no linear relationship
12
Correlation

• Closer to ? 1 the stronger
• Relationships do not necessarily mean what you
  think, i.e. non-causal relationships

13
Spurious Correlation

• Coincidental Correlation chance relationships
• Indirect Correlation related through some third
  variable
• Infant mortality and temperature in country of
  birth are linearly related, but the poorest
  countries are closest to the equator.
• .

14
Putting a value on the Linear Relationship

• Pearsons Product Moment Correlation Coefficient
  (PPM)
• Parametric data - Quantitative data where it can
  be assumed both variables are normally
  distributed, r

15
Putting a value on the Linear Relationship

• Spearmans Rank Correlation Coefficient
• Non parametric - Ordinal data or quantitative
  data where one (or both) variables are not
  normally distributed. Calculated from the ranked
  data ? (rho)

16
Regression

• Identify the nature of the relationship
• Predict one variable from the other
• The independent variable (plotted on the x-axis)
  determines the dependant variable (plotted on the
  y-axis)

17
The Regression line
The Method of Least Squares (the smallest sum of
the squared distances)
18
The Regression equation
Y bX a Y the y-axis value X the x-axis
value b the gradient (slope) of the line a
the intercept point with the y - axis
19
The Regression prediction ?

• Residuals
• Coefficient of Determination
• Coefficient of Determination 100
• R squared (R2)
• How good a fit the equation (and the line) is to
  the data