In the experimental design project, the article was analyzed and reviewed, then its results were presented to other interested students as a seminar. Reliability engineers must determine the lifetime distribution of system elements in order to design and manage resources. Uncertainty information is usually obtained from life tests. In order to obtain accurate conclusions, the analysis of uncertainty data should reflect the test instructions. Traditional uncertainty analyzes usually assume independent observations and assume a Weibull distribution to model the lifetime distribution. However, in many real-world experiments, the observations are correlated, as in cases where the items for the experiments come from different blocks. A reliability engineer has conducted an experiment to investigate the effect of temperature on battery life. Experiments are conducted as a completely randomized block design. There are three temperature levels (15-70-125) and three blocks (three groups of batteries). Each block contains 8 batteries, each of which is placed in 3 rooms (treatment). Also, the second type of tests has been implemented in order to ensure the presence of 4 failures in each treatment-block combination. The presented model includes two random effects: the effect of blocks and the effect of treatments. In this case, the test units are rooms while the observation units are batteries. It is very important to consider the test guidelines and framework in every test plan, otherwise it will lead to wrong results. In this article, NLMM model was presented in order to analyze data reliability with random block and subsample. According to the simulation results, it was found that this model has the least amount of deviation. In addition, the amount of deviation is generally independent of the β parameter value in this method. The traditional method with blocking provides a simple way to calculate the block, but it is weak in determining the sub-samples. If the goal is to check the influence of the factors, the two-step method is suitable, because it keeps the value of the first type error constant at 0.05. The main advantage of the NLMM method over the two-step method is the ability to achieve more accurate results. In general, applying the test protocol will lead to much more reliable results.