Browsing by Author "Dalberg, Cliona Georgia"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Eesti elektritarbimise prognoos
    (Tartu Ülikool, 2015) Dalberg, Cliona Georgia; Kangro, Raul, juhendaja; Lassmann, Joosep, juhendaja; Tartu Ülikool. Matemaatika-informaatikateaduskond; Tartu Ülikool. Matemaatilise statistika instituut
    Käesolevas magistritöös prognoositakse Eesti elektritarbimist 24 tundi ette. Antakse ülevaade tugivektorregressiooni teooriast ning kasutatavast paketist R tarkvaras. Koostatakse ennustamiseks lineaarse regressiooni mudelid ning tehakse nende analoogid tugivektorregressiooni abil. Võrdlemiseks kasutatakse ka ARIMA mudelit. Tulemusi hinnatakse 2015. aasta jaanuari ning veebruari prognooside keskmise suhtelise vea ning keskmise ruutvea põhjal. Mudeleid parandatakse argumenttunnuste lisamise ning muutmisega. Lõplik valik parima mudeli osas tehakse uue testperioodi kaasamisel.
  • Thumbnail Image
    Item
    Imputation of inventories in Estonian Commercial Register
    (Tartu Ülikool, 2013-06-10) Dalberg, Cliona Georgia; Vähi, Mare, juhendaja; Tamm, Ebu, juhendaja; Tartu Ülikool. Matemaatika-informaatikateaduskond; Tartu Ülikool. Matemaatilise statistika instituut
    Missing values is a problem which often troubles statisticians, because most of the analysis methods consider full data. Almost every dataset has unobserved values due to the unconsciousness of the respondents, technical errors and several other reasons. One way to deal with the missingness is trying to replace missing values ‒ imputing. Imputing is a cost-effective measure, it allows to use data which otherwise would be discarded. Imputing also minimizes bias and makes using rectangular dataset and complete data analysis possible (Longford 2005, p. 38; Scheffer 2002, p. 156). This bachelor thesis was written as a part in the project “Integrating annual bookkeeping reports into statistical production system”, which was requested from Statistics Estonia by Eurostat. Data were collected from annual reports of Estonian Commercial Register. Main goal was to complete the section of inventories in the dataset of 2011. The first part of the thesis concentrates on giving overview about missing patterns and applications and theory of selected methods. In the second part a simulation is carried out, the dataset is described and arranged and previously specified methods are tested. Code of the program is added to appendix.