What are the plots I can do in order to select predictors? First thing I do with the data is to check the relation of the variable to be predicted with other variables. To do this I produce simple plots using plot function or qplot function. I do matrix plot and box plot as well. Are there any other types of plots that I can use in order to check choose variables?
 A: You seem to be in the stage that is usually termed exploratory data analysis -- any kind of visualization is potentially useful. My favourite R tools are:


*

*Direct pairwise comparisons with pairs() from graphics and also corrplot() from corrplot package (visualization of correlation matrices). 

*There are huge possibilities for visualization complex data with "facetted" or "trellis" plots in lattice and ggplot2 packages. Extremely effective once you get the idea how these plots work.

*Ordination plots from PCA, RDA, CCA, CA, multidimensional scaling, etc. All of these seek to reduce the dimensionality of your data by seeking the main axes of variation. They have two advantages: (1) They show you the co-variation in a visually cleaner way than pairplots, and (2) the extracted axes can themselves be used as predictors, which may come handy if you have a lot of correlated variables and you are unsure which one should you use.

*3D scatterplots, contourplots, hexbin plots for data with extremely large N (hexbin package), ... there are probably heaps of other kinds of plots. 
