Linear regression modelling


I have a dataset which contains a long list of variables (categorical eg. gender, zipcode group etc and continuous eg. income, home value etc) of responders for a fundraising campaign. I need to develop a model for the “Donation Amount” (in dollars), in terms of a suitable number of independent variables by “Linear Regression”.

How do I adopt an iterative approach (efficient method) of including or excluding variables from the model?

So how do I go about producing these?
a) a parsimonious model with statistically assessed variables.
b) measures of performance and prediction accuracy of the model for the “Donation Amount”, on training as well as validation data.
c) measures of model goodness of fit of the model.
d) details of the results of the statistical tests on the variables that defend the inclusion / exclusion of these variables in the model.
e) Diagnostic plots of the final model presented.

Thanks for help!
Stepwise regression is the way this is typically accomplished. Variables are added (or subtracted) incrementally to the model, then their impact is assessed with a measure like Akaike Information Criterion (AIC). The function in R (my statistical home) is step().