|
Session Title: Variance Explained and Explaining Variance: An Overview of Variance in General, in the General Linear Model, and in Statistical Programs
|
|
Panel Session 618 to be held in D'Alesandro Room on Friday, November 9, 1:55 PM to 3:25 PM
|
|
Sponsored by the Quantitative Methods: Theory and Design TIG
|
| Chair(s): |
| Julius Najab,
George Mason University,
jnajab@gmu.edu
|
| Abstract:
Variance is a crucial concept in common evaluation analytic procedures. Unfortunately, the concept is elusive to many researchers and evaluators. Most statistical texts and courses do not emphasize how different statistical models handle variance. Our aim is to examine and explain variance for evaluators unfamiliar with the concept. These three presentations will describe and explain the relevance and importance of variance. The first presentation covers variance in various distributions. The second presentation applies the basics of variance into the General Linear Model with specific regression, Analysis of Covariance, and Repeated Measures Analysis of Variance examples. The final presentation will examine how different statistical programs utilize variance in different analyses. The discussions of variance are orientated towards a comprehensive description, absent the advanced technical jargon.
|
|
Variance in Distributions
|
| Julius Najab,
George Mason University,
jnajab@gmu.edu
|
|
Variance in the data is rarely the focus of traditional statistical courses and texts. The variability in data distributions is an assumption for every statistical analysis. Researchers frequently assume a normal distribution (the bell curve) in the data. The normal distribution is based on specific data variability or variance and the normal distribution has restrictions to subsequent analyses and inferences we researchers should make. I intend to describe the concept of variance and provide various data distribution examples. By the end of the presentation the non-quantitative evaluators should be able to understand variance conceptually.
|
|
|
Variance Within the General Linear Model
|
| Susan Han,
George Mason University,
shan8@gmu.edu
|
|
Data analytic procedures differ in the way that they allow researchers to partition variance due to an effect or error. The General Linear Model (GLM) is for various univariate and multivariate data analytic procedures. The GLM is an overarching model encompassing multiple regression, Analysis of Covariance and Repeated Measure Analysis of Variance and many others. This presentation described how the GLM utilizes the data variance. Those common procedures in evaluation deal with variance uniquely and those differences are important to the results interpretation.
| |
|
To Choose or Not to Choose: Examining the Generalized Linear Model (GLM) Default Options in R, Statistical Package for the Social Sciences (SPSS), and Statistical Analysis System (SAS)
|
| Caroline Wiley,
University of Arizona,
crhummel@u.arizona.edu
|
|
Understanding how different statistical packages treat variance is pertinent to ensuring that the obtained results reflect what the evaluator ultimately draws inferences about; the inferences drawn ought to match the specifications of the model. However, developing a deeper understanding of the options available will help specify more accurate models. In addition to the multiple methods a specific analysis deals with variance many statistical packages may and probably do treat the same analysis both statistically and conceptually differently. It is therefore important to understand both the default and available options in the package of choice and how the choices one makes or does not make affects the observed outcomes.
| |