This is my first question ever on Stats StackExchange and I am hoping to get some guidance from the community.
I have graduated with a BSc in Statistics, so I am familiar with the fundamentals of Mathematical Statistics and Applied Statistics. I have pretty high grades and think I truly do understand a little bit of Math. Stats (just a little bit). Recently I got my first job as an analyst and I am a bit baffled by how different applying statistical knowledge is relative to college assignments. It feels like a Wild Wild West as I would come up with seemingly decent experimental design to follow with statistical tests only to realize or conjucture conditions that could exist that could violate the required assumptions.
I am very eager to successfully complete the task I am assigned to do at work but, given that's my first job-related statistical analysis ever, I am really nervous if everything is correct.
The task is as follows: my company pays a search engine to highlight its website. Generally, they do it for the brand name and its various versions and for the product name + the brand name. The experiment they want to conduct is to see whether paying a search engine (quite a significant sum) makes a statistical difference relative to the number of sales through the website.
They propose picking several regions of the country and assign them to groups as follows: some regions continue paying a search engine both for the brand name versions and for a bundle the brand name + the product name, some continue paying only for a bundle the brand name + the product name, some stop paying for those search keys completely.
At the first glance, that didn't seem like a complicated statistical analysis at all - I would use ANOVA just like I have done in college many times. However, the more I start to go deeper into the actual procedures and pitfalls, I realize that I, sort of, don't know what to do at all. The company I am working for is an insurance company, therefore, its products are financial in nature. At the very least, my analysis should be taking into account an issue of the Time Series. Additionally, it seems, there is an issue of spatial autocorrelation.
What I am set on doing is the following: I want to pick approximately 20 regions randomly in R conditioned on them being similar in terms of GDP: -7%, +7% interval for that allowed. Of those 20 regions, I want to randomly assign them to the three different groups as described above.
Initially, I thought of doing a simple ANOVA or n-ANOVA analysis, however, I realized that there are two issues I have no idea how to account for: since the sales of a financial product are time series by definition (at least so it seems to me, I don't have a formal financial education and by the time I realized I should take a course in the Time Series Analysis, I ran out of credits to afford it), it has to be adding additional statistical limitations that should be accounted for like seasonality, trends etc. and also I should account for Spatial Autocorrelation.
Could I ask the community to give me some guidance as to what kind of similar cases I could look at, where to find those or point me to some literature of similar topics. Some general guidance is also much welcome!
Thank you!
