JMP Sandwich Estimator
To generate a sandwich estimate of a SE in JMP do the following:
- Run the multiple regression like usual
- Save the leverage plots
- this generates 2 columns for each of your leverage plots
- One column is the Y of the plot the other is the X of
- Check this by doing a simple regression.
- Center the X and Y for the variable you want to estimate
- You will have to create new columns
- Just subtract off the mean
- Now you can compute the beta-hat by averaging the
product of these two columns divided by the average of
the square of the X column. Check this by creating a
product column and the square column averaging and dividing.
- Compute the variance by sum(LX2
- Use the product column you create above, call it P.
- sum(p2) is simple the mean squared plus the
variance of P.
- A simillar argument works for LX.
- Alternatevly you can use a formula: sum(p2)sum(LX2)
Last modified: Thu Apr 19 10:39:30 2001