Dear Nahid, 

non-parametric statistics Spearman rank (=rs)

I think you need to provide more detail. You can't really calculate a correlation coefficient unless it is a 2x2 table or there is some ordering in the 2xm table.

You can calculate a statistic such as Ctamer's V if both variables are nominal, but it isn't very meaningful (in my view) because it is inherently non-directional - which isn't helpful if you wan't to interpret the result or compare between tables.

thank you dear Beatrice, you solved my problem.

thank you Thom, please pardon my confusion for I have learned statistic through my experience with my own research result, but I think if both of my nominal variables have 3 values , it makes a 3*3 table and if less than 20% of the cell have values lower than 5, I can rely on pearson's chi square p value, have I got it wrong?

The chi square will show the statistical significant between the groups, but it does not show the magnitude of correlation..

So,if you are looking for the significant only, chi is an option. 

Please note that the Spearman correlation requires ranked data.  Even if the data is ranked, it would not be recommended for a 3x3 table due to the extremely large number of tied ranks.  Cramer's V as suggested by Mr. Baguley, above, is one choice among measures based on the Chi-Square statistic; however, as he notes, it has its weaknesses. 

Thank you Rae but as I mentioned the chi square is not valid due to low frequency cells more than 20% as well for this reason I can't use Cramer's v

More information would be helpful.  For example, WHY do these cells have low values.  Is it because your overall sample size is small?  Is it because those cells represent very rare occurrences, perhaps it is due to non-random sampling.  When you speak of low cells sizes are the expected values also low?