Challenge: dependent variable observed randomly over time

Hello everyone,

I have been working on this problem and couldn't find a solution so any kind of help is more than welcome!

I have data ranging from the year 1900 to 2000 in 150 cities for four independent variables (one of which is a dummy). These nicely constitute the requirements of a panel; for each variable all observatios are available for each year and city.

The problem however is presented by the dependent variable; for each city I have only a limited number (ranging 2 to 5) of observations which are randomly spread between 1900 and 2000. So for London I have the dependent variable only for 1909 and 1967 while for Paris I have only observations in 1934, 1956 and 1980. (As you've guessed correctly; these numbers are illustrative...) I cannot assume a linear (or otherwise normal) distribution between these observations to 'generate' additional data...

The main question is: is it possible (preferably using Stata) to estimate a relation between the dependent variable (of which I have only a small number of observations) and the independent variables (whereby one is a dummy variable and all observations are present) while retaining the possible effects of both cities and time?

What are your thoughts on this?

I have thought of doing a time series analysis with a dummy for each city, or doing cross-section analysis whereby I group observations in a certain time-frame (e.g. 5-year cohort).

Many thanks in advance! :tup: