I was thinking again about this formula. Is better if you try this other:

Betaj = Bj *(SD(Xj)/SD(Y)) 

The result is the same and I'm sure you have the Standard Desviations (SD), not so the square sums.

Hola Laura.

Es muy fácil si dispones de los datos, te adjunto archivo en Word con las fórmulas y explicaciones.

Si te queda alguna duda o no se entienden las fórmulas hablamos por correo:

Miguel Cañete

[email protected]

I was thinking again about this formula. Is better if you try this other:

Betaj = Bj *(SD(Xj)/SD(Y)) 

The result is the same and I'm sure you have the Standard Desviations (SD), not so the square sums.

If you are using SPSS, it reports standardized beta in the coefficient table just beside the un-standardized coefficients. However, I think you are probably using the GLM option where standardized beta is not reported. There are two ways of doing that, one way is to calculate z-scores (in that case you will need standard errors to calculate standardized version of variables) of both the dependent and independent variables and run the regression on standardized variables; and the other way is to use standard deviations that Miguel just showed above. Hope it helps.  

And the third way to do it is to convert all your measures to z-scores and rerun the analysis with z-scored measures.  SPSS makes this process pretty easy when asking for descriptives (check the box that says "save standardized variables").

Converting all measures to z scores in multilevel modeling is generally not recommended as it could impact standard errors.

May I ask what about Standardized Beta for categorical variables? Can we also apply the formula that Miguel Ángel Cañete suggested above? I think it is meaningless to have mean and SD of categorical variables like gender or education level.

In addition, I did apply two methods (method 1: use the formula above, method 2: standardize all IVs and DV, then run regression model with all standardized variables) but the Standardized B from two methods are quite different. It is correct for continuous (age) IV only.

Thank you very much @Khandoker Mahmudur Rahman, @Julia B. Smith

Regression techniques assume continuous quantitative variables with normal distribution. This is not the appropriate model for dichotomous or categorical variables.