Keywords : Maximum likelihood estimates
Building Discriminant Model For Repeated Measurements Data Under Autoregressive (AR-1) Covariance Structure For patients with diabetes
journal of kerbala university,
Volume 10, Issue 1, Pages 97-112
discriminant analysis is a statistical technique Based on a sample of individuals Taken from communities known in advance, In order to build a model that could help to assigned the group that belong to the new individual. In This Research discriminant analysis used to analysis data from Repeated measurements design, We Will Deal With The Problem of Discrimination And Classification In The Case of Two Groups Under The Assumption of Multivariate Normality For Univariate Repeated Measures Data .
Researchers who studied this problem (Roy & Khattree, 2005), where he presented a descriptive study of the two methods under different structures of the covariance matrix To reduce the number of parameters is required to build a classification rule, While researchers (Kshirsagar & Albert, 1993) studied two methods Growth curve and ANCOVA models for descriptive discriminant analysis To describe the relative importance of the occasions repeated measurements to distinguish between groups.
The importance of this research represented to find the best model to Classify a Group of Patients Who Suffer From Diabetes, For The Purpose of Studying The Effects of The Number of Correlations, Variances, and Number of Repeated Measurements on The Performance of Classification Rules For This Type of Data , Based on Monthly Measurements of Glycosylated Hemoglobin (HbA1C) In The Blood Was Taken In Three Stages, Which Is The Beginning of The Experiment, and After Three Months, and Then Six Months for two groups of patients, the first group consists of (38) patients was Suffered from diabetes type I and the second group includes (33) patients Suffered from diabetes type II,
which has modeled by assuming the Autoregressive (AR-1) covariance structure To reduce the number of parameters is required to build a classification rule Across a Range of Conditions of Homogeneity and Heterogeneity For The Covariance Matrix. In Addition to Assuming Covariance Structures we Will Assume The Structured Mean Vectors Without Time Effect on each Individual. And Some of Computational Schemes For Maximum Likelihood Estimates of Required Population Parameters are Given.
And Through this research, concluded that when the number of parameters began to increase, Thus, the apparent error rate Begin to increasing, And this is what reduces the efficiency of classification rules for this type of data. And We recommend by using the linear discriminant function under (AR-1) Covariance Structures, When you focus on the least number of parameters to build the Classification rule.