Predicting children's height

Using Galton's data to predict children's height using parents height

Sasa Pavkovic

Background information

  • The data are recorded in class intervals of width 1.0 in. He used non-integer values for the center of each class interval because of the strong bias toward integral inches.
  • All of the heights of female children were multiplied by 1.08 before tablulation to compensate for sex differences.

Model

In the data analysis a simple linear model is built.

Using this model we are trying to predict the height of a child in inches.

lm1 <- lm(child~parent,data=galton)
lm1
## 
## Call:
## lm(formula = child ~ parent, data = galton)
## 
## Coefficients:
## (Intercept)       parent  
##      23.942        0.646

Prediction

Prediction is done by using the coefficients of the linear model based on the mid-parent height of parents.

Here is an example with mid-parent height of 75 inches.

newdata <- c(75)
prediction <- lm1$coef[1] + lm1$coef[2] * newdata
prediction
## (Intercept) 
##       72.41

Additional information