4

I have a linear regression model predicting Y from X1 and X2. However, from separate analysis in the past, the independent variables X1 and X2 are known to be accurate a% and b%.

Is there a way to predict Y by accommodating the accuracy rates of the independent variables? Any references (book or articles) you can point is very much appreciated.

Silverfish
  • 20,678
  • 23
  • 92
  • 180
dexter2323
  • 49
  • 1

1 Answers1

1

Assuming that

the independent variables X1 and X2 are known to be accurate a% and b%

means that $X1 = aZ1$ and $X2 = bZ2$ where $Z1$ and $Z2$ are the true values of the independent variables without measurement error, and $a$ and $b$ are known, then you can simply compute $Z1$ and $Z2$ and use those as the predictors in your model.

Robert Long
  • 53,316
  • 10
  • 84
  • 148
  • Thanks for clarifying, Robert. I don't have Z2 and Z2. The a and b are old estimates from a old sample. In the current sample, i only have X1 and X2 and don't know their accuracy rates. However, it is reasonable to assume that the accuracy rates are same. Since the sample is different, i have a aggregate estimate of accuracy rate and I am wondering if i can use these aggregate estimates somehow in the model.... maybe as weights? – dexter2323 Jun 18 '16 at 19:40
  • Even if the samples are different (though presumably from the same population ?) since a and b are the same you just compute Z1= X1/a and likewise for Z2. – Robert Long Jun 19 '16 at 21:59