Unfortunately learning time is directly proportional to power and flexibility of software
Full power and flexibility is only possible when you go past menus to learn the package's command language
Learning the command language allows one to practice reproducible research by being able to re-run entire analysis
Excel is not a statistical package, and it contains severe errors. It also is tedious to use, especially when doing repetitive calculations or graphs
In order of increasing power and learning time: SigmaStat -> SPSS -> SAS -> Stata -> S-Plus -> R
R and S-Plus are superior for exploratory data analysis, graphics, and statistical modeling
Stata is superior for statistical modeling and is good for graphics
R is free to all from http://www.r-project.org; it uses virtually same commands as S-Plus but very limited menus. R runs faster, has slightly fewer bugs, and is easy to install
An R add-on package R Commander (Rcmdr) adds SPSS-like menus to R
There are many excellent online and published texts for learning R, and the Department of Biostatistics holds workshops