Given an interaction how to find the region of var1 at which var2 has significant effect on response Y? I have run a binomial glm in R using proportional data in the form y <- cbind. 
I have two continuous explanatory variables, from which var1 is significant and var2 is not, but the interaction is also significant. So I want to find which are the values of var1 for which var2 results in an effect on y.
I have discovered a function called getRegion(), which I found in two tutorials on the internet, but I haven't found the package that contains it. Are there any other functions that could work in a data set like mine (i.e., proportional response to two continuous explanatory variables)?
 A: 
So I want to find which are the values of var1 for which var2 results
  in an effect on y.

The phrasing of your question suggests a bit that you expect the points contributing to this effect to be in some closed region, however, it is very likely that the effect is simply spread across the entire range of your observations. 
Is there some theory or a previous study suggesting otherwise? If so it should give you an idea what the intervals are...
I am not sure what you are looking for, but I don't think there is a statistical procedure to solve what you are asking for. All 'automated' functions will only try to cut the data into intervals and fit something locally.
If you have no theory, no supporting evidence, but still want to do something because you have a strong hunch, I recommend:


*

*Divide the data into two halves randomly and set one half aside (I
hope you have enough observations, otherwise there is not much point
thinking about 'regions of interaction' in the first place) 

*Fit a LOESS estimator first. It is a non-parametric smoother that fits a curve through your data. Inspect the plots and
look for sudden change of slope, curvature etc. If you find really
convincing evidence of changing behaviour across the range to identify the regions.

*Fit a spline regression and set the 'knots' at the borders of the regions. (or a piecewise linear regression if you prefer.)

*Realize that what you have done now, is effectively i. you looked at your data, and ii. drew a line through it where it seemed convenient. That has nothing to do with statistics. But don't despair! 

*Take the other half of your data and use your model to predict the dependent variable. (You can calculate the MSE, but I would simply plot the actual and predicted values)


If the fit seems obviously wrong, then you have just learned something about the dangers of data mining. If the model fits perfectly... 
Think whether the regions aren't caused by some systematic flaw in the study design, or measurements. And if possible get more data and re-test.
