I don't quite understand what you are doing but here is a link that may help. If not ask again. My research area is prostate cancer, currently. Best wishes, David Booth Link:https://surveillance.cancer.gov/joinpoint/analyses.html

It contains samples and tables that suggest joinpoint selection relative to data points. httpss://surveillance.cancer.gov/help/joinpoint/setting-parameters/method-and-parameters-tab/number-of-joinpoints.

Firstly thank for your answers

In fact, i will perform the regression analysis in order to see if the incidence change over year by calculating the annual percentage change. in the United State's study they have explained that the best fitting model was determined with a minimum of three observations required between two joinpoints, as well as a minimum of three observations required between a joinpoint and either end of the data.

So, my question is: these number of observation have to be calculated before to perform the analysis or we can you use the default number "2". In the link that you have put, they didn't explain how to measure it.

Thank you

Chaima Chebil AHHH Now I understand this is like a loess smooth in some ways. Look at this link, it may be helpful: https://www.bing.com/search?FORM=U528DF&PC=U528&q=How+to+set+up+a+loess+smooth

Best wishes, D. Booth

Chaima Chebil

I am not at ease with the rule of the thumb. One, make sure the assumptions are met. Two, the method you are using, be it grid search , there is justification.

There should be, on average, at least two data points between consecutive joinpoints. If your data points are less than 20(years?), and since your interest is in the change in trend over years (exploring the annual percentage change), use a default of 2 pls. Otherwise, follow the U.S. based researcher's recommendation of [3].

The use of Joinpoint Regression in temporary trend analysis within epidemiology, bio and medical field research is just being explored (if I may say that).

It has several

elements of strength; aims of identifying

time point(s) where the trend significantly changes, so

generating quantitative inferences, rather than qualitative.

Good luck.

David Eugene Booth Abiodun Christian Ibiloye thank you for your answer.