Statistical relationship with both dependent and independent variables percentages that sum to 1 I'm trying to see if there is any relationship between the market share of a company (dependent variable) and the "tweet share" of this company (independent variable). The tweet share of a company is simply the amount of tweets about this company divided by the amount of all tweets about all companies that are in the same industry. So for example if Nike has a 40% tweet share in the sneaker industry this means that 40% of the tweets about sneakers are about Nike. I have data for the first three quarters of 2015.
The problem I'm facing is that I don't know how to check for statistical relationship since both variables are percentages that sum to 1. Any help would be much appreciated.
 A: Since your output lies on $[0,1]$ you should be able to do a simple GLM with support $[0,1]$, the most common being logit or probit links (don't do this actually, the interpretation isn't correct as your independents are also bounded, let me have another think about it). I don't think having dependency inside the independent variable really matters, it is dependency across independent variables which is the problem.
The only problem you might face is determining how many companies you should include, with the problem being due to your constraint, when $n$ becomes large more observations are essentially going to be $(0,0)$ (a lot of small companies have 0 twitter share and 0 market share), which will both greatly skew your coefficients and greatly increase their p-values (as they are all centered around the same value).
A: Maybe a proportional regression model would be appropriate here with an appropriate choice of linke function.  In SPSS Statistics, this can be done with the STATS PROPOR REGR extension command, which can be installed via the Extensions > Extension Hub menu.
