How i know if my multivariate data set is normal distribution? I have  a data set with 4 attributes,i want to know if normal or not.
ex:
attributes(age,height)
calculate standard deviation ,mean for age.
calculate standard deviation ,mean for height.
will be same as calculate age and height as one point,   for checking normal or not?.
If every attributes is normal ,this means that at one point will be also normal?. 
 A: If you're planning on fitting any sort of linear model or a specific case of it (e.g. ANOVA), you only need is for the residuals of your independent variable to be normally distributed. Your independent variables don't need to be normally distributed. Even your dependent variable doesn't need to be normally distributed, as long as the residuals, i.e. the errors left after you've fitted your model, are normally distributed.
If you're using R, the best and easiest way to check for normality of residuals is to plot a histogram of residuals from the fitted model. You can do that easily like so:
your_model <- lm(weight ~ height, data = your_data) #Fit the model

hist(resid(your_model))

If the resulting histogram looks normally distributed, then your assumption of normality should be met. If it does not look normally distributed, there may be other variables accounting for a lot of the variation that you have not included in the model (e.g. in your case, possibly gender). Or if the distribution is skewed, you may need to fit a non-normal error distribution to your data, e.g. lognormal (although that shouldn't be the case with height and weight data). 
There are also mathematical tests for checking the assumption of normality, such as the Kolmogorov-Smirnov test or the Shapiro-Wilk test. These are also implemented in R, but I've seen quite a few statistical papers advising against their use.
