Time series/Regression question. I am trying to figure out what variables most significantly affect the download rate of my app. Right now I have a data frame with a column for dates, downloads per day, payouts per day, and a few more other variables that correspond to the date column. 
This may be impossible but I'm looking for some way to be able to predict how many downloads I will have based on the values of the factors I have. 
 A: What you can do is find the correlation of the features to find which are more correlated and eliminate any feature (not recommended) which has very low correlation. Or you can apply random forest to find the feature importance and then apply linear regression to find the number of downloads. 
A: This task can be down by variable selection methods. I suggest dynamic regression and autocorrelated residuals (DREGAR). To explain more, the model is like,
\begin{align}
& y_t=\sum_{i=1}^{p} \phi_i y_{t-i} + x'_t \beta +\epsilon_t\\
& \epsilon_t = \sum_{j=1}^{q}\theta_j\epsilon_{t-j} +e_t
\end{align}
where $e_t\sim \,i.i.d \,\,N(0,\sigma^2)$ and $x$ can contain many covariates. Now if you use an $l_1$ penalized likelihood to estimating the parameters, you get a set of (selected) covariates in $x$ and  some (selected) orders for AR term. This is a time-dependent variable selection method. Fortunately, it is implemented in the R package DREGAR See here that you can use it. 
Update: from your comment, you have 1600 points that is enough for removing bias in this model.
A: Have you tried building an ARIMA model, incorporating the additional variables as leading indicators?  In this exercise you would build a model that fits your purposes/criteria and optimizes over AIC (my preference) or SIC/BIC.  You can then use the model to forecast expected downloads per day.  Assuming you do use lagged versions of the other variables you have (referenced above as leading indicators), you will be able to quantify the degree to which these variables drive the downloads.
All of this is easily doable in R.  I would look in to the 'forecast' package.
