Metrics and covariates are structures to organize non-timeseries and non-MRI measures. Examples include surveys, BP measures, and polysomnography scoring.
Covariates
A "covariate" is a stored value. The covariate has a name, and will be either a number or text ("string"). It will usually have a date (and possibly time) when it was measured.
The covariate information is stored in the "covariate" table. The actual values are stored separately as values in the "covariate_value" table.
Covariate table
| Read only | |||||
| 'covariate_id' | 'int(11)' | 'NO' | 'PRI' | '' | 'auto_increment' |
| 'modified' | 'timestamp' | 'NO' | '' | 'CURRENT_TIMESTAMP' | 'on update CURRENT_TIMESTAMP' |
| Required | |||||
| 'name' | 'varchar(255)' | 'NO' | '' | '' | '' |
| 'valtype' | 'enum(''number'',''string'')' | 'YES' | '' | '' | '' |
| Optional | |||||
| 'valdefault' | 'varchar(255)' | 'YES' | '' | '' | '' |
| 'category' | 'varchar(255)' | 'YES' | '' | '' | '' |
| 'description' | 'varchar(4096)' | 'YES' | '' | '' | '' |
| 'parent_covariate_id' | 'int(11)' | 'YES' | '' | '' | '' |
| 'formula' | 'varchar(2048)' | 'YES' | '' | '' | '' |
Covariate_value table
| Read only | |||||
| 'covariate_value_id' | 'int(11)' | 'NO' | 'PRI' | '' | 'auto_increment' |
| 'modified' | 'timestamp' | 'NO' | '' | 'CURRENT_TIMESTAMP' | 'on update CURRENT_TIMESTAMP' |
| Required | |||||
| 'subject_id' | 'int(11)' | 'NO' | '' | '' | '' |
| 'covariate_id' | 'int(11)' | 'NO' | '' | '' | '' |
| 'value' | 'varchar(255)' | 'YES' | '' | '' | '' |
| Optional | |||||
| 'name' | 'varchar(255)' | 'YES' | '' | '' | '' |
| 'datetime' | 'datetime' | 'YES' | '' | '' | '' |
| 'projectgroupstudy_id' | 'int(11)' | 'YES' | '' | '' | '' |
| 'analysissetgroupstudy_id' | 'int(11)' | 'YES' | '' | '' | '' |
| 'tag' | 'varchar(1024)' | 'YES' | '' | '' | '' |
| 'comment' | 'varchar(2048)' | 'YES' | '' | '' | '' |
Metrics
A metric is a way to group several covariates. For example, if you do a survey with several questionnaires, each questionnaire can be saved as a covariate value, but if you have a lot of questionnaires, it will become unwieldy. A metric could be "MRI visit survey" and consist of the questions and questionnaires asked on the day of the MRI visit. You can then link this to a study in a project. For example, to link polysomnography scores to a study with MRI's and physiology, create a new study "metric" and then in the project link that to the study, just like you link an MRI series or physiology recording.
A "metric" is just a type of study associated with the subject, set in the subjects GUI. A separate very simple table records all covariate_values in that metric.
Protocol ID for metric
You can use the same format of protocol id for the metric as for MRI and physiology; because studies are selected by type, they will not get confused.
\[patient id][session number][yyyymmdd]
Table metric
| 'study_id' | 'int(11)' | 'NO' |
| 'covariate_value_id' | 'int(11)' | 'NO' |
Note that one metric could have repeated entries - this might be useful to make up for missing data (I'm not sure). And of course one metric can be used in multiple project studies.