It is easy to load data in as spreadsheet or any statistical software with a friendly user interface. It has also become relatively easy to browse through the menus and fit some kind of regression, plot a few graphs, get some p-values and hurray!! You are a hero! But wait are the results valid? is the model appropriate for the type of data and the question you are interested in answering? Is there an objective in the first place? (Are you looking for a suitable problem to go with your brilliant solution!?) Technological advancement has increased our statistical computing capabilities but let us be careful not to confuse our ability to use these tools with our understanding of the subject that is statistics. It is perilous; it like having ones head in an oven and feet in ice and so that one can feel warm. The good thing is that from our experience in offering statistics help a lot of misuses are not intentional, just ignorance. It is critical that you know when to seek statistics help, know your abilities and their limits.
Some of the ways in which people may abuse statistics include: bad data collection, improper application of statistical functions and formation the wrong conclusion from the results. A historical example of a wrong conclusion was the case where the incidence of the rickets was strongly correlated to being in certain families. Thus, early scientists concluded that rickets was hereditary. It was later documented that rickets was the result of malnutrition, and that poverty, not rickets, was inherited. People get confounded, it always a good idea to share your thoughts on your data and results with peers and whenever possible seek statistics help or input from someone more experienced than you are.
A few frequently misinterpreted things that we have to mention for now:
Correlation does not imply causation. You need more than just correlation, if A causes B, then A should at least occur before B occurs, there should be some dose-response relationship between them. There are many more criteria for causation that we don’t intend to discuss for now. The point is correlation does not imply causation.
“Statistically significant” does not necessarily mean important. Does, for example, a statistically significant increase of 0.00005mmHg in the average systolic blood pressure in a population matter?. Equality important is that “Not statistically significant” is not the same as zero or no effect.
It is appropriate to seek statistics help whenever you feel inadequate, actually even if you are a seasoned data analyst and understand most (no one is a geek in everything) of the principles a second eye is always as good or crucial as your next breath.