There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
| Principal Investigator | |
| Principal Investigator's Name: | Joseph Graves |
| Institution: | North Carolina Agricultural and Technical State Unviersity |
| Department: | Applied Science and Technology - Data Science |
| Country: | |
| Proposed Analysis: | A precision matrix is commonly used to uncover undirected network structures in a wide range of applications including genomics, biomedicine, and finance. The precision matrix estimation remains less addressed for multimodal missing data. We propose a direct sparse estimation method for the sample precision matrix without imputation when high dimensional multiple modal data contain block missing data. Our method is motivated by the DISCOM method for linear prediction using multimodal missing data (Yu et al., 2020). Our proposed method is achieved by three steps. First, we compute a sample covariance matrix using all available data without deleting or imputing missing data. Second, we derive a modified matrix using the computed covariance matrix to guarantee the existence of a precision matrix. Last, we incorporate a graphical LASSO technique to construct a sparse precision matrix. Our proposed method is compared to other precision matrices estimated using complete data only, low rank imputation, and MICE/MIFA imputation. Simulation studies show the effectiveness of our proposed method over these compared methods in terms of the Frobenius norm, edge-sensitivity, and edge-specificity. This study demonstrates that our direct estimation method of precision matrix overcomes a decreased sample size in the complete data and uncertainty of imputation. As our simulation study shows promising results, we plan to compare our proposed method to other methods using real applicaton data. Since numerous articles have used the ADNI data for the multimodal data research, we would like to access the ADNI database with the same reason. We expect our method to help uncover hidden networks among multimodal neuroimage data such as CT, PET, and MRI in a different way. We also want to compare multiple precision matrice for the different categories of the Mini Mental State Examination (MMSE). Our project was initiated to enhance the research capability at an HBCU under NSF Grant #2100729. Yu, G., Li, Q., Shen, D., Liu, Y.(2020). Optimal sparse linear prediction for block-missing multi-modality data without imputation. Journal of the American Statistical Association, 115(531), 1406-1419. |
| Additional Investigators | |
| Investigator's Name: | Seongtae Kim |
| Proposed Analysis: | A precision matrix is commonly used to uncover undirected network structures in a wide range of applications including genomics, biomedicine, and finance. The precision matrix estimation remains less addressed for multimodal missing data. We propose a direct sparse estimation method for the sample precision matrix without imputation when high dimensional multiple modal data contain block missing data. Our method is motivated by the DISCOM method for linear prediction using multimodal missing data (Yu et al., 2020). Our proposed method is achieved by three steps. First, we compute a sample covariance matrix using all available data without deleting or imputing missing data. Second, we derive a modified matrix using the computed covariance matrix to guarantee the existence of a precision matrix. Last, we incorporate a graphical LASSO technique to construct a sparse precision matrix. Our proposed method is compared to other precision matrices estimated using complete data only, low rank imputation, and MICE/MIFA imputation. Simulation studies show the effectiveness of our proposed method over these compared methods in terms of the Frobenius norm, edge-sensitivity, and edge-specificity. This study demonstrates that our direct estimation method of precision matrix overcomes a decreased sample size in the complete data and uncertainty of imputation. As our simulation study shows promising results, we plan to compare our proposed method to other methods using real application data. Since numerous articles have used the ADNI data for the multimodal data research, we would like to access the ADNI database with the same reason. We expect our method to help uncover hidden networks among multimodal neuroimage data such as CT, PET, and MRI in a different way. We also want to compare multiple precision matrices for the different categories of the Mini Mental State Examination (MMSE). Our project was initiated to enhance the research capability at an HBCU under NSF Grant #2100729. Yu, G., Li, Q., Shen, D., Liu, Y.(2020). Optimal sparse linear prediction for block-missing multi-modality data without imputation. Journal of the American Statistical Association, 115(531), 1406-1419. |
