Panel data analysis

[Sonia Zam]

Hi everyone,

Need help with analysing this data. A description of the data: Individuals are biannually (or annually if they are high risk) invited to a screening program until the age of 60. After the age of 60, individuals can participate into the program but through self-referral. When individuals participate in the program some demographic and other information is collected during each participation round. I am interested in knowing what factors are associated with continuing to participate in the program (self-referral) after reaching the age limit of 60. Any suggestion what is the appropriate statistical model to analyse this type of data? Thanks in advance for your help

Below is a sample of the data. Thanks.

id age participation_date smoking familyhistorycancer gender

B001 58 5-Nov-12 0 1 0

B001 60 7-May-15 1 1 0

B001 62 9-Jun-17 0 1 0

B002 57 28-Nov-11 0 0 0

B002 58 22-Oct-12 0 0 0

B002 59 25-Nov-13 1 1 0

B002 60 10-Nov-14 1 1 0

B002 61 14-Dec-15 0 1 0

B002 62 12-Dec-16 0 1 0

B003 63 25-Feb-12 0 0 1

B003 64 8-Oct-13 0 0 1

B003 67 12-Feb-16 0 0 1

B004 66 16-Jun-12 1 0 0

B004 68 22-May-14 1 0 0

B005 59 17-Nov-12 1 1 1

B006 59 12-Mar-13 1 0 1

B006 61 29-Apr-15 0 0 1

B007 58 22-Jun-12 0 1 0

B007 60 4-Aug-14 0 1 0

B008 61 29-Jan-13 0 0 1

B008 63 4-Mar-15 0 1 1

B008 65 30-Mar-17 0 1 1

B009 59 28-Jan-15 1 0 0

B010 58 28-Feb-12 0 0 0

B010 61 3-Jun-14 0 0 0

B011 63 4-Feb-13 1 0 1

B011 66 17-Sep-15 1 0 1

• Delete
• Move to

[Phil Bromiley]

You will increase your chances of useful answer by following the FAQ on asking questions-provide Stata code in code delimiters, readable Stata output, and sample data using dataex. While you provide data, to use that one would have to do the work of getting it into Stata. It would increase the chances of someone helping you by make it easy and using dataex for example.

I suspect that the answer depends on how you want to model or think of continuing after 60. If you just want a 0/1 that they continued or not, then I suspect you are looking at a cross-sectional logit or something similar. If you only have one observation that you're trying to explain per individual, then the panel analysis techniques really don't work very well. If you are concerned with how soon they continued than a survival analysis might be appropriate. If you are concerned with how many times they continued than a panel Poisson might be appropriate.

[Carlo Lazzaro]

Sonia.
in my opinion a substantive issue is whether undergoing screening after 60 years is funded by health care system or entirely out-of-pocket by patient and/or her/his family. If the latter were the case, you might end up with many zeros that only reflect patient's economic constraints (that, as such, should be separated from the zeros that arise from patient's decision to stop screening).