|Aggregation in OM refers to the bringing together of data for the purposes of statistical analysis. It could be data from one client, across a number of clients, or across a number of services. |
The purpose of aggregating data is often to get larger data sets that give greater statistical power and precision of estimation. However, if dissimilar data are aggregated together findings from the aggregated dataset may not accurately represent any of the smaller datasets so thinking through what will be lost and what will be gained by aggregating data is really important and it is reassuring in reports to see some comment on similarity of aggregated groups. A typical example in therapy data is aggregation across gender. Clearly combining all gender categories for a service gives a complete picture and the larger n will give more precision. However, if there marked differences in what is being analysed, e.g. change scores, aggregation across gender will be concealing potentially useful information.
As ever: Quality of data in = quality of data out; however, poor quality thinking can ruin even good quality data!
|Typical aggregation of OM data starts across clients but generally adds aggregation across therapists and across services and across periods of time. Subgroup analyses can test to see if there are marked dissimilarities between the datasets being aggregated, e.g. do the different therapists have markedly different distributions of clients’ starting OM scores.|
Multilevel modelling accounts the problems that arise when aggregated data are “nested”, e.g. OM data may come from multiple sessions from the same client: so occasion is said to be nested within client. Multiple clients will be nested within therapists and therapists within services.
Try also … #
Nothing here yet