Hi,

I have a data from a survey done in 2005 and followed up in 2011. In 2005, 500 participants were involved in the survey. In 2011, only 90 participants agreed to participate. There is a huge loss to follow-up numbers. My question is:

- would it be statistically reliable if I combine 2005 and 2011 data to make a pooled data and get effect estimates? would those effect estimates be reliable keeping in view the number of loss to followup participants?

- if I want to compare the results of 2005 data to 2011, is it statistically reliable to do as the number of participants was reduced enough in 2011?

out of these two options, which one is more statistically reliable to publish?