Construction of optimal investment portfolio is very complicated task due to many diverse factors which might affect risk and return of the portfolio in the future. Values and impact level of unique factors on the portfolio are changing over time; therefore every investor should take into account the fact that there always will be a certain level of risk associated with any portfolio involving stocks. There is a number of ways to form a collection of most appropriate stocks and bonds for investment portfolio as well as to allocate weights of assets based on various criteria. All of these methods, dedicated for selection and allocation of assets, have their specific features and some disadvantages. In order to be able to conclude which of the asset selection methods have least disadvantages, four popular techniques were analysed and compared. These techniques were based on different variables: correlation coefficients between asset returns, maximisation of the utility function (diverse values of risk aversion coefficients were analysed), selection of assets with highest historical returns, and employment of modified price-to-earnings ratio. The article deals with multistage extension to the mean-variance and expected utility maximisation portfolio choice. Multistage investing consists of several essential stages, where each stage forms a basis for the next stage by providing useful input data, derived by stage-specific analysis. For construction of optimal portfolio the following stages are used: asset allocation, security selection, investment strategy development, construction of the model and its evaluation. After asset allocation was made and stocks for the portfolios had been selected, different theoretical asset allocation models (equal weight asset allocation, Markowitz model, Capital Asset Pricing Model (CAPM), model where risk free asset is incorporated when constructing a portfolio) have been modified and adapted in order to become suitable for real market situation. Such prerequisites as normal distribution of stock returns were not satisfied by most Lithuanian companies’ stocks when different interims were investigated, therefore authors set a presupposition that distribution properties of the stocks can be disregarded when Markowitz and CAPM models were applied to real market. Some other changes for the prerequisites of models were made; otherwise these theoretical models could not have been applied to Lithuanian market. After all models had been applied in Lithuanian equity market, back testing was carried out and certain characteristics of outcomes of different investment strategies were compared. Results were judged against characteristics of popular stock exchange indices of Baltic States in order to obtain conclusions. Most models were developed for broad financials markets (global markets). In the paper we analyse financial market of Lithuania. Since this market does not fit assumed conditions of general models, the models were slightly modified to apply for Lithuania market. The results of portfolio were compared with Baltic States index. It was concluded that the highest return rate is achieved by constructing the investment portfolio with employing modified Capital Asset Pricing Model. The best technique for selecting stocks proved to be the maximisation of utility function when risk aversion coefficient A=3. In addition to this, after comparison of different asset selection methods, it was noted that the highest value of the Sharpe ratio was achieved by utilising the same technique. After investigation it was noted that investors should add a risk free asset into portfolio of stocks because it usually improves the results of most portfolios, irrespective of their contents. Constructing portfolios based on asset allocation according to indices analysed in the paper is not recommended because characteristics of indices were worse than the ones of constructed portfolios. Stocks of every company quoted in NASDAQ OMX Baltic (2011) Stock Exchange in Vilnius Official list for more than 10 years (2001 beginning – 2010 end) were investigated. Stocks of 14 companies satisfied preset 10 year interim criterion.DOI: http://dx.doi.org/10.5755/j01.ee.23.2.1542