I have unstandardized regression coefficients and standard errors from a multilevel regression model.
So I guess I have to transform first the standard errors to standard deviations in order to use the formula to transform unstandardized to standardized.
beta = b (sd(j)/sd(i))
(where sd(j) is the standard deviation of the independent variable and sd(i) is the standard deviation of the dependent variable).
However, I know I am not getting correct values because I have one sample of beta values already obtained by the authors themselves and they do not match. So, I guess the rest are not the true values I'm looking for.
This is the formula I am using for transforming standard errors to standard deviations:
SD = SE * sqrt(N)
If you have any thoughts on this, please let me know. Maybe I am using not apporpiate formulas for these types of regression models.
Best,
David
Standardize your variables and run the regression model again will work. David Booth
If you don't have anything more than a regression coefficient and it's standard error I'm afraid you are out of luck unless the author that published this is willing to help you. Sorry, David Booth
As it turns out, the first formula was wrong.
Here is the corrected version:
beta = b (sd(i)/sd(j))
(where sd(j) is the standard deviation of the independent variable and sd(i) is the standard deviation of the dependent variable).
As you can see, the standard deviation of the independent variable goes into the denominator. I double checked it by comparing with my sample of beta regression coefficients provided by the authors and they all match.
So, question solved. Thanks for the help.
I need to compute the standardized regression coefficients for a power analysis and have:
The SE of the regression coefficients
The unstandardized regression coefficient (b) and the t value
The sample size (n)
Is it possible to compute the standardized regression coefficients (beta)?
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression