I am able to do time trend analysis (incidence rate) for the individual countries but can't seem to be able to do that for both countries. The data am using is age adjusted rates provided by a cancer registry. Thank you.
Joint point programme attempts to fit more than one regression lines to data. Accordingly it gives the estimates of APCs. Since the regression lines fitted to different countries may vary by time segments it is not possible to compare them. However, it is better to compare there latest APC estimates and see whether they vary to a great extent or tally with each other. Another way is to compare the pooled estimates of APC for both the countries which is given in the output. To explain with one example: let us say that you are comparing two countries A and B and data considered is for years 1982-2011. For country A, the join point shows three time segments namely 1982-89; 1990-1997 and 1998-2011. For country B, the time segments are 1982-87; 1988-1993, 1993-2005 and 2006-2011. Obviously, it is difficult to compare APCs because at no point of time they correspond to same period. In such a situation, go for pooled estimates of APCs provided by the software and compare whether they differ. Another way is compare the APCs of 2006-11 of country B with that of 1998-2011 of country A. Both are assume to suggest the latest consistent trend. Then decide whether country A is better or worse compare to country A. Both the option are logical. But when you talk of pooled estimates your time period is too long. It is often better to give weightage to latest trend ( Country A-1998-2011 and Country B= 2006-11) and compare.
Combine the data from two countries together, define a new dummy variable, country, depending what model you use, say if you use Proc Reg in SAS, need country=0 or 1 , but if you use Proc GLM you can use the country name as the value. But you need use class statement.
You can use the same model, you used but add two parameters
country and country*time.
country*time is the interaction of the time and country, which shows the trend difference between two countries. If it is significant, it means that the trends for two countries are different.
With interaction, the time effect parameter is only for the country, which is used as reference group. for other one, you need add time parameter+Contry*time parameter together to get the trend for that no-ref country.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude