If TB infection is  binary then is your dependent variable then you can use logistic regression

Dear Adjuik

Yes I have data like binary. Thank you very much for suggestion.

Regards

Gehendra Mahara

PhD 

If you have access to geographically coded data, you could also do a cluster analysis to identify the geographical locations spatially (with high TB rates; high PM2.5 levels). If your data is for several time periods, you can add a spatio-temporal analysis.

Use Poisson Regression to adjust TB rates for PM2.5 levels.

Good luck with your project.

Dear Admin

Thank you very much for your suggestion.

Thanks

Gehendra Mahara

 Dear Gahendra

Your aim of the study and the methodology used are not clear from the given description. Need more details about your work to help you to select appropriate statistical tools

Dear Jayadevan

Thank you for your concern toward my question. Actually, You know that it has been established that Beijing is the most vulnerable place in the world, in the context of Air pollution. And there still have not been studied about PM2.5 and its association with Tuberculosis infection. Thus, I want see it the association between it and also with meteorological parameters as well.

I look forward your answer.

Thanks

Gehendra Mahara

PhD Research Scholar

China has cut their TB rate by 50%.  See recent report at BBC.

http://www.bbc.com/news/health-26611859

Dear Prof Dr Amin

Thank you for your reference. Probably I will go through Spatial-Temporal analysis by using Poisson Regression analysis or might be Bayesian Analysis.

Thank You again.

With Regards

Gehendra

If you will use a non-Bayesian approach, consider using SaTScan. It is free, and it is used in over 100 countries for disease surveillance.

Ref: www.satscan.org

it depends on the hypothesis you testing, if you looking about PM2.5 is associated or increasing risk of TB infection, and TB is binary variable, then you can use Logistic regression .

and in case if you time between the PM2.5 exposure until the time they got TB, then it's to use COX regression and represent it through KM curve as well.