CODEWITHSUNDEEP
r
sumalatha
sumalatha

Reputation: 79

Perform basic regression

I want to know how to do the following:

 a <- data.frame(num = 1:10, numsqr = (1:10)^2)
 b <- data.frame(num = 11:14, numsqr = 0)
 fit <- lm(numsqr ~ num, data = a)

 b$numsqr <- predict(fit, b)

 print(b)   

   num numsqr
 1  11    121
 2  12    144
 3  13    169
 4  14    196

Right now I'm getting the following result

  print(b) 
    num numsqr
  1  11     99
  2  12    110
  3  13    121
  4  14    132

How could I get my anticipated result??

Upvotes: 0

Views: 54

Answers (2)

Anant Gupta
Anant Gupta

Reputation: 1149

If you assume that you have no idea whatsoever of the relationship of the data ( in this case you know that y= x^2 ), it will be very difficult to get the exact values through a linear regression

You can try converting the response variable to a logarithm for a slightly better resolution to incorporate the curvature

Upvotes: 0

Rorschach
Rorschach

Reputation: 32426

To get the squared variable in the formula, you can use I or poly (still linear in the coefficients), otherwise it is just fitting y ~ ax + b.

fit <- lm(numsqr ~ I(num^2), data=a)
fit <- lm(numsqr ~ poly(num, 2), data=a)  # different model, same predictions

predict(fit, newdata=b)
#   1   2   3   4 
# 121 144 169 196

Upvotes: 3

Related Questions