Time Series model for discrete sparse data

[michellekwai]

I have a set of dataset for time and value:

Jan 100; 

Feb .;

March 200;

April .;

May .;

...; 

I impute missing value as 0, but the dataset is very sparse for TSA model. Any good suggestion for missing value treatment and also how to make this kind of dataset good for TSA(eg, ARIMA, ARIMAX) to forecast future values? 

Thank you so much!

[Reeza]

How many data points do you have in total, and how many missing?

[michellekwai]

724 observations and 2 variables for month and value. some months have value, some are not. Since these values are about loan amount every month, so I just treat all missing value as numeric zero. 

so dataset will be like (100,0,0,0,0,200,0,0,0....) , sparse dataset. Not sure dataset like his is good for ARIMA or other TSA model? 

[Reeza]

How many are missing from the 724? Your approach will vary based on the amount missing. 

If you check PROC TIMESERIES it has some options on how you can fill in your missing data, or you can use a multiple imputation approach or some other option....context matters.