3 sites are sampled pre-tx, and the same 3 sites post-tx. I am trying to compare the number of cocci in site 1,2,3 pre and post tx (refer to table) But the issue is i've recorded my values in ranges (i.e. categories of <10; 10-20;20-50;50-100 & >100). What statistical test should I be using, or is there any i can use with this?
Recording continuous or ordinal values only in bins is a serious design flaw that will not be remedied by any statistical method. You can discarded a great deal of information/power/precision.
If site represents your independent samples, your sample size is 3 so you do not have power to do a statistical test (binning makes the effective sample size even lower than 3).
Note that pre-post analyses require a very stringent assumption that nothing else changed between pre and post. For next time you might consider a randomized crossover design.
As answered, the data is not in a form that lends itself to answer the question at hand. If, however, nothing else is available, perhaps you might try to fit a loglinear model to the three way contingency table Site x Visit x Cat (where Cat is a nominal variable with levels <10, 10-20, etc.)
Doing that, if the effect of Visit is significant, you wold have a lead that something changed between visits, and looking the percentages of observations in different levels of Cat you would probably have a good idea of the direction of change.
Another idea would be to cross-classiffy Pre- and Post- Cat and test for symmetry. Essentially, this means that a subject e.g. in the pre- 10-20 level of Cat is as likely to end in a post- e.g. 20-50 as one initially in 20-50 to end in 10-20. I do not know if that would make sense in your setting.
