The starting point for any analysis should be the research question rather than the dataset. Having a well defined hypothesis will help in framing the analysis and in deciding which source to use and which analysis methods are appropriate. However attractive, resist the temptation to decide on the source first and to then trawl or dredge for "interesting" analysis - this rarely leads to interesting or useful insights. Your research question will also guide you as to which countries you wish to include in your analysis. For example if your research question is about the performance of educational systems you will want to make sure you include countries that have different educational structures. Similarly, if your hypothesis is something to on social or political attitudes you will want to include countries that have different political systems, structures and traditions.

One pitfall that many researchers fall in to is to over complicate the question and to lose sight of the key question. It is often a good discipline to try and articulate the key question in 2 or 3 sentences. For some reason, explaining to some one verbally why it is an interesting topic or question helps crystallise what needs to be done. Imagine responding to a comment of "so what". It is very easy to lose sight of the simplicity of the question in preparing research proposals or in synthesising the theory.