Multi-Dimensional Analysis is an Informational Analysis on data which takes into account many different relationships, each of which represents a dimension. For example, a retail analyst may want to understand the relationships among sales by region, by quarter, by demographic distribution (income, education level, gender), by product. Multi-dimensional analysis will yield results for these complex relationships.
Multi-Dimensional Analysis is generally used in statistics, econometrics and other related fields and the results of this kind of analysis used in the different fields can be further applied to different fields like business enterprise. Multi-dimensional analysis actually is a process which groups data into two basic categories which are the data dimension category and the measurement category. To illustrate this, let us take the case of a football game.
A data set which consists of the number of wins for one football team every year for many years could be categorized into a single dimensional or longitudinal data set. Another data set which consists of the number of wins many different football teams within a year can be under a single dimensional or cross sectional data set. A single data set that consists of the number of wins for various football teams across many years could be contained in a two-dimensional data set.
Two dimensional data sets are also called panel data in other disciplines. Logically, any two or higher dimensional data sets could actually be considered as multidimensional data but the term multidimensional data tends to be applied on data sets only with three or more dimensions.
For instance, there are data sets used for forecasting which provide forecasts for various target periods and these are carried out by multiple forecasters made at multiple horizons. All three dimensions can provide for better information which can be gleaned from two dimensional panel data sets.
In a multidimensional the term dimension refers to a structural attribute of a data cube. The dimension is composed or related and hierarchical members. For instance, the "Time" dimension may have the members like years, quarters, months, weeks, day, hour and so on. In the same manner, the "Geography" dimension may have members like regions, countries, cities and so on.
A dimension member is an element of any given dimension just like in the example above where year like years, quarters, months and weeks are members of the "Time" dimension.
There is also a dimension hierarchy is a way to organize dimension members into parent and child relationships. In the "Time" dimension example, the month is the child belonging to the quarter which in turn is the child to a year.
A dimension title refers to the name used to make the dimension known. In the above examples, the "Time" and "Geography are dimension titles.
The dimension title member is the name of the member as in the case of month or city. The dimension value member is an instance of a dimension member. For example, 2007 is the value of the dimension value which is Year.
A data point refers to the intersection of multiple dimensions while a data value resides at the data point.
Multidimensional analysis is very important in a business enterprise because they are the basis for some of the decisions of the business organization which will give them better edge over the competitor. Today's business environment is constantly evolving and business trends change very fast so it is always a good idea to analyze enterprise related things.
Multidimensional analysis uses dimensions and measures to analyse data. To do this using Business Intelligence tools such as Cognos 8 you usually need to first model your data dimensionally.
Dimensions are hierarchies and have one or more levels. What dimensions you define depends on your data and your business model. The user can look at the data at any level – for example Team level will show totals of all the members of that team, at Area level you will see the total for all teams within that area.
Measures are usually quantities such as quantity sold, total revenue and so on. Once a measure is selected in the analysis the measure is aggregated to the level you are analysing at. So if you are analysing at Area level
and you have selected the total revenue as your measure, you would see the total revenue, aggregated to Area level.
Often the data you require will be organised into a data warehouse consisting of dimensional and fact tables and produced by a skilled data warehouse team.
The data is then modelled as metadata and published to make it available for analysis. In Cognos 8 this is achieved using the Framework Manager application. Hierarchies for each dimension are stored in the model so do not need to be defined by the user.
OLAP Data Sources
Sometimes the data source itself is dimensional, for example a SAP B/W cube or a Cognos Powercube. These data sources can be input directly into the Framework model and do not require further modelling.