Home  |  Call Us: 562-244-0845 | info@flexopartners.com

FPi Statistical Consulting Services offers clients a one-stop place for study design, sampling schemes, and statistical analysis and modeling.

Study Design: FPi will develop or review existing study parameters to assure information that is collected meets an identified sent of goals and objectives, and cost parameters. Specific design issues that we have investigated include appropriate data collection methodologies, sample size, questionnaire design, question scales, screening criteria, questionnaire length, incidences and costs, among money other design issues. We are also experts on issues surrounding sampling balancing and data weighting having conducted thousands of research studies.

Sampling Design: Which is the best sampling approach and what sample size would be optimal, balancing margin of error and costs? There are typically numerous sampling procedures that can be employed to collected public opinion and marketing research information. Random sampling, stratified sampling, and cluster sampling are just a few. Selecting the most approached sampling design will assure that the sampling data is projectable to the population of interest at the lowest possible error and cost.

Different sample sizes produce different margins of error. The calculations of sample sizes for different sampling designs requires different formulas and choices. We can do the mathematics, and help design sample size options that maximize accuracy and minimize costs.  

Statistical Analysis and Modeling: Once data is collected it needs to be analyzed in ways that produce actionable conclusions. At FPi, we have extensive, wide-ranging experience in standard and advanced analytical procedures, and can handle any size statistical problem. Examples of our analysis and modeling capabilities are described below.

• Cross-Tabulations: This is standard output in a public opinion and marketing research study. It includes frequency and % responses to all questionnaire items, and tables of the responses of key subgroups. We can typically produce a set of standard cross-tabulation tables in less than one day. Also outputted are standard data formats for more advanced analysis, including Excel spreadsheets, SPSS and SAS data formats.

• Correlation Analysis and Linear Regression: Correlation analysis measures the strength of a linear association between two question responses or variables. In Linear Regression, the goal is to establish a correlation relationship expressed in an equation. The equation can then be used to predict the outcome of one variable, given changes in another variable. When only one variable is used in the prediction, it is called simple linear regression, and when more than one is used, it is called multiple regression.  

• Parametric and Non Parametric Statistical Testing: There are many mathematical tests that can be run to determine how related variables are to each other. The most appropriate test is selected based on a number of factors about the data type. Parametric statistical tests like t-tests and F-tests are used when the data is normally distributed. However, often data is not "normally" distributed. In this case a special class of statistical tests, called nonparametric tests, must be used. We will work with you to find the most appropriate test of statistical significance given the circumstances of the data sets.

• Data Reduction: Often you need to make sense of the patterns in the relationships between variables in large data sets, or to reduce large data sets into more manageable subsets, dimensions or factors. Doing this requires techniques that identify the underlying statistical relationships between variables. Data reduction techniques greatly simplify the description of large data sets, and often help uncover latent associations between variables and responses.

• Classifying Respondents: Statistical procedures exist for grouping or classifying the way people respond to questions. We statistically model how the responses of people are alike, and how they are different. And, based on these response patterns, we develop unique subgroups of people, each with their own opinions, behaviors and demographic characteristics. This is an excellent way to identify different segments in the population, and those that offer the best targets for your product or service.

• Time Series Analysis: The purpose of time series analysis is to see how things change over a specified period. For instance, you might want to test the affect of an advertising campaign over a period of time, or compare the effects over time of one marketing mix against another. We can measure and manage any kind of time series data that you might have available. In the analysis, we can provide diagnostic assessments, and build forecast models that will help you estimate the effects of future actions based on the patterns uncovered in actions you have taken in the past.

• Structural Equation Modeling (SEM): The goal of a structural equation modeling assignment is to help clients determine the extent to which different theoretical models define a data set. The models can describe simple associations between variables (such as in regression or path analysis models), or more complex associations and interactions between known variables and unknown or latent constructs (as in confirmatory factor models). Our job is to define the structure of different models, and test them, to see which offers the best solution. Once we have identified the optimal model, the associations and interactions are estimated to help understand how the variables are related and which are critically important and need attention.  

© 2005 Scott W. Flexo, Ph.D. and Flexo & Partners, Inc. All Rights Reserved