摘要：Straipsnyje supažindinama su aktyvumo indeksų taikymu ekonominiams procesams analizuoti ir prognozuoti užsienio šalyse, pateikiama Lietuvos realaus ekonomikos sektoriaus indekso LBIX R sudarymo ir skaičiavimo metodika. Pagrindinis straipsnio tikslas - parodyti, kad aktyvumo indeksai įtraukti į prognozuoti bendrąjį vidaus produktą (BVP) skirtus modelius, leidžia patobulinti prognozavimo algoritmus ir patikslinti BVP prognozes. Aktyvumo indeksų naudojimo būtinumą lemia ir tai, kad jiems skaičiuoti reikiami duomenys gaunami gerokai anksčiau negu skelbiamas BVP.
其他摘要：In this article the Lithuanian real sector activity index LBIX R is presented. This index is calculated without using weights and values, i.e. all data arc provided in physical tenus. That means inflationary processes arc eliminated. Time series data arc aggregated into 5 sub-indices: industry, agriculture, transport, communication and construction. The activity indices were calculated for all of these sectors, and after that they were combined into the consolidated index LBIX R. Similar indices arc successfully used at many central banks, e.g. Japan, Germany, France, Russia, etc. The methodology of calculation of LBIX R is also presented.The second part of this paper is devoted to describe different models of GDP. The join behaviour of Lithuanian GDP, exports of goods and services and average monthly salaries is examined by the structural vector auto-regression models (SVAR). Striving for the larger accuracy, apart the aggregated indicators their components arc analysed as well. To describe relations of cointegration, vector error correction model (VECM) was used. The accuracy of forecasts that were calculated as a sum of forecasts of separate parts of GDP was made more precise by using the residual of cointegration.In the third part of this paper different models of GDP were added by activity indices as new regressors. An accuracy of the forecast of each model differs. Therefore, two kind. of errors were calculated: ordinary errors a. difference between actual data and forecasted values and Jack knife errors (in absolute terms or modulus). Two periods of modelling results were analysed: before the Russian crisis and after it. Such general conclusion can be made: before the Russian crisis better forecasts of GDP were got using VEC model, and after the Russian crisis VEC model with LBIX R index as a new regressor gave better results, i.e. the average modulus errors of this model were the least.
