ASR provides independent statistical services for clients using data sets derived from a variety of sources.

Sampling

Sampling is the statistical process to determine how many units of information are required to reliably represent the population being surveyed. ASR can determine sample size to produce statistically sound information. Statistical confidence in the survey results depends on how well the data represents the population surveyed. Too small a sample makes the results questionable. But equally, too large a sample is wasteful and unnecessarily expensive. The best balance can be achieved by good sample design. In addition to producing a statistically valid sample, we can help you determine the most cost effective sample for your survey needs - whether they are simple, single population opinion surveys, or complex, multi-faceted surveys.

Advanced statistical analysis

ASR has the expertise and tools to produce sophisticated multivariate analysis and statistical modeling of data sets. These procedures include factor analysis, discriminant analysis and general linear model multivariate analysis. ASR can also provide on-line statistical reporting and graphical presentation of data. This service is particularly valuable if you are routinely collecting data from multiple sources and want to keep users up to date with results on-line rather than through expensive printed reports. Statistical analyses can be performed on-line into an existing database and the results presented in tabular or graphical format using a specially built driver that interfaces with SuperWEB from Space-Time Research.

Validation

ASR can design data validation routines at data input or output stages. Data input is usually controlled through software mechanisms. However, this may involve using different collection methods as well. Data output validation involves statistical tests to ensure that data falls within pre-defined limits or benchmarks. This may involve transformation of scores into benchmarks or other formats to meet reporting criteria.

Data cleansing and mining

Before using data, it must be tested to ensure that it is valid, reliable and contains no spurious information. Web-collected data can be controlled within our Survey Manager software. But where automatic data validation is not possible, for example open ended questions, or text-based responses where a respondent can enter either a percentage value or a numerical value, it is important to clean data and provide a consistent set of responses. Given the sophisticated databases used to hold survey data and the volume of data that can be captured and held, manual data cleansing can be daunting.

Data mining techniques can be applied to complex data sets to automate the extraction of hidden predictive information contained in these data. Data mining is an extension of statistical analysis and provides solutions to business decision problems. For example, data mining is a particularly useful when applied to customer relationship management (CRM) data sets. It can efficiently uncover trends and allow users to build customer behavior models. These, in turn, can be used to predict customer response to a new product.

ASR can assist customers in the design of data mining explorations into their own data sets. These are particularly useful where successive sets of information are built over time or through repeat surveys.