I have Y variable and X variable. R2 is 0,002, when I add seasonal dummy R2 increased to 0,01. However it is the only 2-3 variables X to add. Which method of regression can I use to predict more accurately? Or should I use kind of regression trees`?... The main task is to get see whether initial X1 influence Y and what is the effect. The graph below shows the fitted line when i regress X1 on Y. May be I should some how play with the data (i tried to took ln(), but I again model predicts poorly)
You have an enormous amount of data, but no immediately visible structure. You could try spline transformations for your percentage cover to account for possible nonlinearities, but it doesn't really look like this is very promising.
Whatever you do, do a sensitivity analysis by removing the three very high data points and checking whether your results hold up without them. Especially for a spline transformation, they may start to become influential.
Sometimes there simply is nothing there. This may be helpful.
