View
214
Download
0
Embed Size (px)
Citation preview
LINEAR REGRESSION: On to Predictions!
Or: “How to amaze your friends and baffle your
enemies”
Imagine:The correlation
between 1 mile run time and VO2 Max is r = -1.0!
Suppose you know... that a 6:30 mile = 40 ml/kg/minand 6:00 mile = 45 ml/kg/minCould you predict what VO2 Max
a person would have who could run a 5:00 min mile???
Of Course you could, by finding the equivalent point on the line!
Mile Time vs. VO2 Max
MILE TIME
VO2MAX
4:00 4:30 5:00 5:30 6:00 6:30
40
45
50
55
60r = -1.0
Describing and Defining the LINETo describe a line on a
graph, we need to know:The slope The point where the line
intercepts the y axis
More Math?? Y=a+bX
General formula for a straight line
Calculated from the means, s, r
Y = a + b XY = the predicted value of y for a
given value of Xa = the point of the y interceptb = the slope of the line (rise over
run)X = the value of X ( Height) for
predicting Y (Shoe size)
Linear Regression
Maybe you recognize this general equation: Y = a+bX:
VO2 = 111.33 - (0.42 * HR)
Y = a + (-b * X)Y = dependent variableX = independent variable
How Accurate is the Prediction?When the correlation coefficient is
equal to 1.0, then every actual score will fall exactly on the prediction line.
THERE IS NO “ERROR” BETWEEN THE ESTIMATED PREDICTION and REALITY
Mile Time vs. VO2 Max
MILE TIME
VO2MAX
4:00 4:30 5:00 5:30 6:00 6:30
40
45
50
55
60r = -1.0
Get Real! In the REAL
WORLD, it is never so tidy
There is some deviation between the line and most points
Standard Error of EstimateThe predicted (estimated) score will not
be exact, there will be a margin of error between predicted and actual scores.
Thus we need to know the standard deviation of the prediction error.The SEE gives one a feel for the
accuracy of a prediction
Note the error from predicted?
R = .87
Prediction Line
Actual Scores
Body Composition data: Compared to UWW Skinfold 7 site Skinfold 3 site BIA Infrared Circumference
r=.87 SEE= 3.5% r= .87 SEE = 3.5% r=.80 SEE = 5.0% r= .80 SEE= 4.5% r= .75 SEE = 7.0%
UWW vs dissection: SEE = 2.0%
Let’s give it a try!Lab # 4: Predicting Shoe size
(dependent variable - Y)From Height (independent
variable - X)First derive the linear regression
equation, then try it out!