Compared to point estimate, interval estimate of compressive strength of concrete from limited test data would give a better confidence on the evaluated statistics, which can be obtained with Normal approximation theory. In this study, a non-parametric methodology based on bootstrap re-sampling for interval estimate was found to be more suitable. Optimal numbers of bootstrap samples and optimal number of data in each bootstrap sample in such analysis demand attention. Using cube test results, these aspects of bootstrap re-sampling as applied to interval estimate from concrete test data are investigated in this paper. It is recommended that for evaluation of mean or standard deviation of concrete compressive strength, optimum number of bootstrap samples should be between 1,000 and 2,000 with equal to or more than 25 data in each sample. The corresponding numbers for estimation of characteristic strength of concrete was advocated as 4,000 to 5,000 bootstrap samples, each of 30 or more data. Normal approximation theory might yield slightly higher estimate of characteristic strength, which could be detrimental in case of health evaluation or safety margin assessment of important existing structures.