I’m confused about the differences between Two way anova and Regression with dummy variables.
On spss,
I can conduct two way anova through the
general linear model → univariate and
we can do the regression
through Regression → Linear
what are the differences between them?
They both are based on the very same (linear) model, they only focus on different aspects in the analysis of the model. Only some details of interim calculations are different for practical purposes.
Regression focusses on the estimation of coefficients (which can be tested) and on prediction. ANOVA focusses on the impact of predictors in explaining the variance of the response.
General linear models can combine both kinds of predictors (categorical and continuous) and both apsects in one and the same model, what can be used as a mixture of ANOVA and regression analyses. If you focus on the AVONA analysis, you interpret the continuous predictors in the model as covariables and call it "ANCOVA", if you focus on estimating model coefficients you interpret the categorical variables as covariables and call it "regression, adjusted for covariables". You can toss it and turn it whatever way you want. The underlying model is one and the same.
Note that general linear models are usually just called linear models (LMs). GLMs are a further generalization of LMs (= generalized linear models) that allow to analyse response variables that do not need to have a conditional normal distribution (options are: Poisson, exponential, gamma, binomial, negative binomial and others).
Basically , ANOVA interprets the interaction between two categorical independent variables on the dependent variable whereas regression in confined to the relationship between one dependent and categorical (dummy variable) which can later be used as quantitative.
They both are based on the very same (linear) model, they only focus on different aspects in the analysis of the model. Only some details of interim calculations are different for practical purposes.
Regression focusses on the estimation of coefficients (which can be tested) and on prediction. ANOVA focusses on the impact of predictors in explaining the variance of the response.
General linear models can combine both kinds of predictors (categorical and continuous) and both apsects in one and the same model, what can be used as a mixture of ANOVA and regression analyses. If you focus on the AVONA analysis, you interpret the continuous predictors in the model as covariables and call it "ANCOVA", if you focus on estimating model coefficients you interpret the categorical variables as covariables and call it "regression, adjusted for covariables". You can toss it and turn it whatever way you want. The underlying model is one and the same.
Note that general linear models are usually just called linear models (LMs). GLMs are a further generalization of LMs (= generalized linear models) that allow to analyse response variables that do not need to have a conditional normal distribution (options are: Poisson, exponential, gamma, binomial, negative binomial and others).
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression