I have a poverty dataset and I wish to forecast the poverty percentage of a place in the future. I only have 2 years = 2006 and 2012 data. Is time series possible? Is there a statistical way that I could include more data points in between?
I don't think time series is possible for several reasons. First you generally need at least 50 points of data with no interruption in between for time series. Second, if structural breaks occur between the two years how would you test for this?
You could simulate data in between, but you would have to have a reasonable way to guess at what the data is, at least the distribution. That does not sound like a good idea to me....
Other threats, are you don't know if their was a bi-annual spike or other anomalies in the gap years, Also, if you see a trend in 2006, what happens if you assume it continues until 2012, does it match up with those at the beginning of 2012. The huge gap sets off alarms. Do you hypothesize any moving averages across time?
For clarification, are you trying to predict into 2017, using these data? More alarms going off. Have there been any historic changes (bias) that may affect poverty? I am guessing so.
For a fun exercise, what happens if you use 2006 data to predict future data, does it come close to 2012, as mentioned. I am not saying you should seriously do this, but it may be an entertaining activity for novelty purposes!