Hey! Student trying to graduate

#1
Hey everyone,

My name is Charlotte and I am currently an Information Management master student. I am only 5 weeks away from my master thesis deadline. The subject of my master thesis is cross-device and cross-channel conversion and i am writing it at a leading e-commerce company (Amazon-like webshop) where I have set up an experiment in Google AdWords. The idea is that when people are going to buy a product, they orient on their smartphone and make the actual purchase on a larger device. Current attribution models assign conversion to the last (paid) channel used before placing an order, thus assigning too little value to the mobile channel if my hypothesis is correct.

The experiment I have set up is a geo-experiment with four regions. In 3 region I boost a certain type of keyword on mobile devices (there are 3 types) and on 1 region I do nothing. The experiment runs for 4 periods of two weeks. After each period the treatments rotate, so that each region has had the same treatment at some point in time after 4 periods. Why a geo experiment? Because once people put their smartphone in their pocket and switch to a different device, we cannot track them anymore. With a geo experiment I am assuming that most people stay within a certain perimeter.

I am having a hard time figuring out how to do a clean analysis of the data. At the university I have only been taught to work with SPSS and I've been taught the basics of R. With R i've learned how to build classification models, using decision trees, naive bayes, neural networks, logit regression, etc. My thesis coach has urged me to use Stata, which I installed today. At this point I don't know how to start. I don't know how Stata works, I dont know what my dataset must look like, what to aggregate and what not, I don't know which type of analysis I must use to start correcting for time trends and regional effects. With only 5 weeks left until my deadline and high expectations from my employer, I am starting to get nervous. That's why I joined this forum, so I can get some help when needed. The people around me either don't understand that the heck I am doing or they are already statistics experts and go way too fast for me.

My thesis coach has told me to google Fixed Effects, so that's what I am currently trying to figure out :eek:

So that's my story!
 
#2
Hi,
I am sorry i do not know the answer to your stats related question but I have to start writing my master thesis and can REALLY use your help to gradate.
I joined a digital advertising company (performance based outcome pricing) to write a master thesis on cross device and cross platform attribution. They want me to research innovative models that exist and suggest possible partners.
Except, like you said, cross device attribution cannot be tracked! What about probablisitc and deterministic modelling? What is it that i can do? Any topic ideas would be GREATLY appreciated.
Thank you very much, greatly looking forward to some help.