It is very depending in your refinement quality. If you have done a very good refinement with chi2 value around 1, also Bragg R-factor and Rf- Factor values less than 2  .. then the lattice parameter error would be in the fourth decimal digit of Angstrom( err = 0.0001 - 0.0003 ‎Å )  and your certainty could not goes more than that (less than the mentioned number) according to many facts (  the x-ray wave length limit, specimen displacement, surface structure of the material,  the grain size ... etc) which will affect the x-ray measurement.

So I guess 3 decimal places for lattice parameter would be enough 

and you can apply this to density calculation 

any how, you can just search on the web about " uncertainty estimation of measurement"

this one would be helpful for you , if you want to include some error calculation formulas and techniques 

http://bit.ly/1SoorAW

Regards

@ N. Ray: I thing this is not really correct. You do not consider that your peak position is not defined in terms of your step size but by a curvature described by many single measurements which enables a positioning which can be better than that. On the other hand, not only Delta theta but also theta is important. The higher theta the higher the accuracy. But the most important fact is that you do not only use one peak for lattice determination but many. The higher the number of reflections the higher accuracy and precision of the lattice parameters derived. Typically, step widths are in the scale of 0.02° in 2 theta. And this is quite reliable for a lattice parameter determination with an error of +-0.001 Angstrom (or even better as mentioned above). More problematic are the quality of peaks at higher Bragg angles and misalignment effects (zero shift of the detector, misalignment of the sample), absorption uncertainties, inhomogeneous grain sizes etc. But they are somehow (indirectly) considered by the refinement parameters mentioned by Y. Maswadeh as well.  

What about density and volume they are also relaible upto 4 digit after decimal or if step size is less than 0.02.?

Well, this not only depends on the volume bit also on the mass. Only of you know the composition very exact rho=m/V. Have a look at this presentation. All software programs using in their fundamental version the given equations. The scaling and the example of Ettringite should answer your question. For numerous phases, e.g. mixed crystals, unknown crystal water or OH groups, or even as for ettringite the not described  H may change the mass dramatically and so the density.

Conference Paper Simulation of powder diffraction patterns in PowderCell

From your refinement results, you will find the error/uncertainty of lattice parameters, as well as the volume of the cells, i believe. The mass can be slightly different if you have refined occupancy of atoms. Using rho=m/V and the propagation of uncertainty (https://en.wikipedia.org/wiki/Propagation_of_uncertainty), the significant digits of your final rho can be determined.