Modelos mistos em estudo de isquemia cerebral

Abstract

The modeling of data from longitudinal studies stands out in the current scientific scenario, especially in the areas of Health and Life Sciences, since it is common to evaluate the same experimental unit at different times, which induces a correlation between measurements. Thus, the modeling of the intra unit dependency sample is required through the choice of a covariance structure that is able to receive and accommodate the variability in data. However, the lack of methodology for correlated data analysis may result in getting erroneous results, which can lead to erroneous conclusions, and it can also increase the occurrence of Type I error and underestimate the standard errors of the model estimates. In this study, we adopted a Gaussian mixed model for variable response latency ( the time it takes for the animals to find the real hideout) of an experiment investigating the memory deficits in animals subjected to cerebral ischemia when treated with fish oil, the structure of covariance for the random effects used was the unstructured one. To estimate the parameters of the model, we used the methods of maximum likelihood and restricted maximum likelihood. Based on the likelihood ratio test, information criteria and adjustments of models we adopted autoregressive covariance matrix for the errors. The residual statistics analysis and diagnosis for the model were satisfactory, since the obtained results corroborate with biological evidence, that is, it was found the effectiveness of treatment to alleviate the cognitive effects caused by cerebral ischemia