Regression Vs ANOVA???

lickleMo

Hi there.

I'm looking for a bit of direction regarding the use of regression and/or ANOVA. The truth is I'm slightly confused....I understand that we use ANOVAs for categorical data and regression for continuous data....but I've read somewhere that they are the same??

I want to look at the predictors of intelligence i.e. socioeconomic status (high, medium, low), parental education (none, primary school, secondary school, college, other) and length of time breastfed (none, months, 3-6 months, >6 months). I've done up ANOVAs to see if there is significant relationships between these factors and intelligence....I wonder now can I put these into a multiple regression model to see how well they predict intelligence as a whole? Could I then control for socioeconomic status and length of time breastfed to see how much parental education alone contributes to the model using a hierachial multiple regression?

Any help would be greatly appreciated! I'm going goggle-eyed looking at stats books and not getting anywhere!

Thanks!

Victor

I think you need to keep it simple, just do a multiple regression of the data first and see what it throws up.

Parental education alone could be a simple linear regression.

Can you post up a small sample of the data?

lickleMo

Hi Victor.

Thanks for that .

Here is a small sample of the data set:

[Intelligence] [Mothers ed] [Fathers ed] [Socioeconomic] [Time bfed]
[120] [2] [2] [-4.69] [1]
[102] [5] [5] [5.34] [1]
[116] [4] [2] [-5.12] [2.4]
[95] [1] [1] [3.32] [1]
[62] [1] [1] [6.79] [1]

Intelligence and Socioeconomic status are both continuous variables, the rest categorical.

I really appreciate your help!

Thanks again!

2Scoops

ANOVA and regression are conceptually the same statistic. They both compare systematic to non-systematic variance, the only real difference being that the ANOVA uses categorical grouping variables rather than continuous. This does not mean you must use ANOVA, though - you can certainly include your variables in a multiple regression model.

lickleMo

Thanks....so I can put both categorical and continuous variables into a multiple regression?? That's the part that I'm just not 100% sure of!

Jokesetal

It looks like like you're going down the MANOVA route.
Simple PCA (principal components analysis) may help - you look at the eigenvector that gives you the most information and that's your primary discriminator (age, group, history etc.). It gives nice graphis too, plenty of plots where you can circle groupings.
Other techniques that are easy (less than ten clicks using SPSS, Minitab etc.) are DFA and Cluster analysis.