Data Imputation for correlated longitudinal cohort data

[Win Khaing]

Can anyone help me with advice on imputation of data in logitudinal (correlated) data?

In my database, income is my study factor. Assuming Missing at random, income has some missing data in next 20 years follow up (5 years interval of follow up).

1995 - 2000 - 2005 - 2010 - 2015

Income at the baseline measurement, can be varied with follow up (income may increase, income may decrease).

One income in 1995 may increase in another 5 years. 

My data have 20 years follow up. 

So, which imput method is the best for my data?

Thanks in advance.

Win