A basic reference for all styles of science: the Stanford Encyclopedia of Philosophy. A key concept to talk about scientific development is Thomas Kuhn's paradigm, to which Freeman Dyson (2012) added that technology can be more important than ideas (Science 338: 1426-1427). For an actual research project, one of the most important issues is the type of question, ranging from descriptive to predictive, which is often mis-handled and leads to fruitless debates, say Leek and Peng (2015, Science 347: 1314-1315); see also Hofman et al. (2017 Science 355: 486: 488). Subsequently, decisions on the choice of variables, data collection, data cleaning and modeling are much more important than p-values about which there is much more debate, say the same authors (2015, Nature 520: 612).
There are trainings for those who go to dangerous regions.
Most PhD students have a Windows or Mac operating system, write in Word and compute in SPSS or Stata. But don't follow the crowd! It has no wisdom, certainly not on software! Save your files and then replace your Windows system by Linux (e.g. Ubuntu or Mint), write your papers beautifully in LaTeX, and do all your computations, statistics, GIS, and simulations in one of the two major open source languages R and Python. Much better, faster, and cooler! And it's all for free.
For R, which has most to offer to social scientists, there is a user friendly general introduction and a slightly more advanced introduction and a website where you can choose tutorials for geeks or non-geeks. Very handy is being able to take subsets from larger data. An excellent and reasonably comprehensive textbook on statistics with R is: John Fox and Sanford Weisberg (2011, 2nd ed.) An R Companion to Applied Regression. For most specialized topics you'll find tutorials or manuals on the Web, e.g. econometric methods such as panel models and time series. What you will most likely need is information on how to import various sorts of data files (e.g. from SPSS, or other) into R. Also qualitative analysis of texts and webpages can be done. There are also several GIS packages for R (see, among others, this youtube clip), and Google Maps can be used in R as well. Python is very popular among big data scholars, even though big data can also be analyzed in R, and in the natural sciences.
If you need specific methodological advice for your research proposal, you can contact us (via Jeroen Bruggeman).