# ANOVA multi-variable problem

#### prasun samir

##### New Member
If there is a dependent variable depending on 5 independent variables or factors, how to find the relative effect of these 5 groups on dependent variable using ANOVA?
Like what is the most important group affecting dependent variable? We can get the idea using F-statistics but I want to know this dependency in percentage (%) for these individual 5 groups on independent variable.

#### mmercker

##### New Member
Hi, if these independet variables are continuous, you can standardize them before calculating your ANOVA (centring around zero and divided by the standard deviation) and afterwards the effect sizes / regression coefficient are comparable between predictors - they tell you how much the outcome variable changes for 1 sd change in each predictor. Based on this, I guess you can directly calculate some kind of "percentage influence" by setting the sum of the regression coefficients to 100 %. If you have only categorical predictors, you don't have to standardize but the definition of regression coefficients changes (they now tell you how much the outcome variable changes for changing the cateories of the predictor). That means of your predictors have the same number of categories, you can easily compare them. If you have a mix of continuous and categorical variables, I am actually not sure if it makes sense to compare them at all, since they are of different nature...

#### hlsmith

##### Omega Contributor
You can also calculate the r-square. Next you can calculate the partial eta-squared for each variable. This latter value tells you the proportion of r-squared explain for the individual values. You can convert this into a proportion.