Consider two phases show up in FullProf. The intensity of a peak from phase 1 is lets say 1000 counts, and the intensity of another peak from phase 2 is lets say 250 counts. Can I say that: "whole structure consists %25 percent of impurity?" by comparing the intensities of phases?
Better than the reflex intensities are the scale factors, because they are calculated for the whole pattern and not for single reflex intensities. This is an article which has a formula:
another software for Rietveld analysis freely available is MAUD. The author also added a video tutorial to help users interested in quantitative analysis:
I don't know your background, but I'd suggest to read some documentation about XRPD and Rietveld analysis in general before using MAUD. It's kinda user friendly, and you'll always get some numbers out of it, but it'd be important to know: 1) if/how much you can trust those numbers. 2) how to try to make those numbers the closest to what it is in realty in your sample.
Just keep in mind that learning how to do a decent Rietveld analysis does not happen overnight. :)
Perhaps it would be enough for rough but quick estimation to calculate so (your example): 250/(1000+250) = 20%. However, the methods suggested above (WPPD and Rietveld) are more accurate although they require a fairly thorough experiment and subsequent calculations. It all depends on what scientific problem you're going to solve this.
Starting with a %1 parasitic phase is the best as you mentioned. In addition I will try to figure out how whole pattern fits. So I can compare without huge mistakes.
I agree with others who have suggested Rietveld refinement. This will give you the most accurate and quantitative results. There are many free programs available, as have been mentioned. I like to use GSAS. There are free tutorials with data sets that take you step by step through the refinement process for getting phase fractions. Since the data sets are already known, you can quickly determine if you are performing the refinement correctly. Then you should be ready to try it on your own data.