In: Statistics and Probability
Linear Regression
When we use a least-squares line to predict y values for x values beyond the range of x values found in the data, are we extrapolating or interpolating? Are there any concerns about such predictions?
Answer:
Given that,
When we use a least-squares line to predict y values for X values beyond the range of x-values found in the data, are we extrapolating or interpolating? Are there any concerns about such predictions:
From the given information, we can write Yes, there are serious concerns about such a prediction.
Let us consider,
The data of seal weight (kg), overhead width (cm).
So,
The dependent variable (y)=Weight (kg)
The independent predictor variable (x)=Overhead width(cm)
x | 7.9 | 7.6 | 9.3 | 7.1 | 8.9 | 7.4 |
y | 179 | 195 | 266 | 141 | 243 | 184 |
Above table,
The least-square line is,
................(1)
The regression equation y on x,
Now assuming,
The overhead width is to predict the weight of the seal.
The x-value =1.7 cm
"x" value substituting in equation (1), we get
Therefore,
Then,
The predicted weight of seal if overhead weight 1.7(cm) is -96.306.
Beyond the range of x is -96.306.
It is unrealistic.
Since the weight of the seal is never negative.
Predicting results are very concerns.
Note:
Predict the least-square line only in the valid range of regression out of regressor if we predict least square line.
So,
This is a concern.