I have unstandadized B coeff.s in my parameter estimates for ANOVA. How can I use Beta=B(SDx/SDy) to convert them into standardized B's (Beta's).

Why doesn't SPSS do this for me, like it does in regression?


  • $\begingroup$ Furthermore, there is really no analytical advantage to using the standardized coefficients. $\endgroup$ – JKP Feb 2 '12 at 13:53

First, in regression procedure SPSS (as many other software) actually first computes beta, because algorithm works with correlation matrix. It then computes b from beta by the formula you show (if you revert it).

Second, in ANOVA or general linear model procedure we usually deal with categorical predictors (factors) which are internally recoded into dummy or some other type of contrast variables. For such predictors, b is more valuable than beta because b is directly interpretable as difference in means (e.g. between specific groups). This is probably the reason why beta is not printed out.

| cite | improve this answer | |
  • $\begingroup$ Honestly I don't understand you. And you seem to try asking some new question unrelated to your current one. Maybe you overhaul it and post as another question. $\endgroup$ – ttnphns Feb 4 '12 at 8:06
  • $\begingroup$ Sorry for the confusion, too many things going on, apparently in my head. I update this: For parameter estimates for interaction effect in ANOVA, SPSS gives one B value and says the rest is redundant. My question was: how I can transform this B into Beta when standard deviations are not clearly given. $\endgroup$ – JonBonJovi Feb 5 '12 at 12:24
  • $\begingroup$ Why bother head with ANOVA. Want betas? Go to regression. Everything you get with ANOVA you can get via regression, although you might have to make some recodings by hand first (e.g. creating dummies out of categorical variables). $\endgroup$ – ttnphns Feb 5 '12 at 14:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.