DataLiteracy| DataLiteracy | |
| Home Help Backlinks Search Index Recent Changes | History |
Headings: !! Information literacy !! Data Analysis Literacy !! On-line Resources for Learning and Teaching Statistics
The UCLA Information Literacy Initiative seeks to teach students how to "locate, evaluate and use information effectively and ethically." In particular, see Bruin Success with Less Stress. This is an online tutorial designed to help students learn about citation, documentation, intellectual property and academic integrity.
The Texas Information Literacy Tutorial (TILT) is designed to introduce students to research sources and skills, including the use of the Internet and library databases. A module also helps students learn how to critically evaluate the credibility of sources. The tutorial offers an introduction and three modules which take approximatetely 30 minutes each to complete.
Variable Names - conventions of cause and effect
In order to make distinctions between cause and effect, data analysts have adopted the following conventions. Causes are labeled as explanatory or independent variables. Effects are labeled as the dependent or response variables. Causes are also referred to as x-variables while effects are y-variables. This is because in bivariate graphs, like scatter plots, causes are assigned to the abscissa or x-axis while effects are assigned to the ordinate or y-axis.
Reading and using tables
J. Scott Long's tutorial
presents a concise but comprehensive overview of how to read, analyze, test and present data in contingency tables (i.e. the cross-tabulation or cross-classification of two or more variables). It includes discussion on how to make causal inferences (p.5) and how to implement statistical control in multi- or K-way tables (p.9). All examples are done in the Stata analysis package and present the corresponding commands.
Reading and using graphs
Graphical displays of data are used to capture and communicate the properties of variables (i.e. the central tendency, spread, and shape of its distribution), the differences between variables (i.e. multiple y's or multiple x's) and the relationship between variables (i.e. x's and y's). Four basic types are the box plot, histogram, stem and leaf and scatter plot.
The first three are typically used to examine the distribution of individual variables or to compare differences between variables. The fourth, the scatter plot, is typically used to examine relationships among variables. Scatter plots are particularly useful in assessing the magnitude (i.e. slope), and strength (i.e. linearity) of relationships. They are also the workhorse for diagnostic analysis.
A bit more advanced are thematic maps created with GIS software. These use geographic, political and other features (e.g. zip codes, census tracts etc.) to display the spatial distribution of variables (e.g. income, race, water consumption etc.). See our links for more details.
We also have more information on a graphical approach to data known as Exploratory Data Analysis (EDA) and on some principles of graphical design that will help you avoid the "chart junk" that come from the default settings of programs like Excel, which undermine your ability to communicate information clearly.
Significance and the problem of inference
Even if an analysis uncovers a difference or relationship, two interrelated questions loom over the head of any data analyst. The first is whether the observed differences or relations are meaningful. Generally speaking, there are three non-exclusive criteria of significance: statistical, substantive, and causal. The second question is whether they are "real" or spurious. This is the problem of inference: whenever we try to identify the "true" state of the world based on a limited set of observations, there will always be a degree of uncertainty about our conclusions.