Time series data form inputs to the ESPON Database, at spatial scales including NUTS0, NUTS1, NUTS2 and NUTS3. Series are often incomplete at the lower levels in the NUTS hierarchy. The task is to impute the missing values, and ensure the spatial coherence of the estimates. A methodology is presented for data imputation based on an autoregressive model which is fitted to the existing data, and used to impute the missing values, using a Bayesian approach. The spatial coherence of the results is also ensured. The metholodgy has been implemented using the R language, and tested on typical short-run time series for NUTS0, NUTS1, NUTS2 and NUTS3 regions. R and JAGS code to operationalise the methodology is presented in this report, and a suitable workflow is outlined.