# SDA 3.4 Documentation for UPDATES

## NAME

updates - Changes to SDA since Version 3.3

## DESCRIPTION

This is a summary of the differences between Version 3.4 and Version 3.3 of SDA.

## VARIABLE NAMES UP TO 32 CHARACTERS

• SDA variable names may now be up to 32 characters long. (The previous limit was 16 characters.)

## CHANGES TO ANALYSIS PROGRAMS

• TABLES
The procedure to calculate confidence intervals in complex samples has been revised. For percentages in some of the cells of the table, the degrees of freedom used to compute confidence intervals have been reduced, resulting in slightly wider (more conservative) confidence intervals. For more details on how SDA calculates standard errors and confidence intervals in complex samples, see the document on Methods for Computing Standard Errors.

The chi-square probability for complex samples has been replaced by the Rao-Scott probability. The chi-square statistics are still reported when the statistics package is requested. However, the probability values based on those statistic are deceptively low for complex samples. F-statistics based on the first-order Rao-Scott adjusted chi- squares and their associated probabilities are therefore included in the statistics package. For details see the online help file.

The user can choose either the weighted or the unweighted N of cases to display. The default remains the same as before -- namely, to display the N of cases used to calculate the percentages. However, the user can now specify that both the weighted and the unweighted N of cases be displayed, or that neither N be displayed.

Number of decimals can be set for all statistics. There is a set of lines at the bottom of the option form that allows the user to specify the desired number of decimals to display for all of the available statistics.

• MEANS
The procedure to calculate standard errors and confidence intervals in complex samples has been revised. For means and totals in some of the cells of the table, the degrees of freedom used to compute confidence intervals have been reduced, resulting in slightly wider (more conservative) confidence intervals. Also, for differences between means and a reference row or column, the covariance arising from cluster sampling has been included in the calculation of standard errors. Usually such a covariance would be positive, resulting in a reduction in the size of the standard errors and confidence intervals of the differences. For more details on how SDA calculates standard errors and confidence intervals in complex samples, see the document on Methods for Computing Standard Errors.

An option has been added to display the p-value of each difference from the cells in a base row or column. This p-value is based on the difference between each cell mean and the means in the reference row or column. It is a test of the hypothesis that the difference is zero in the population from which the sample was drawn. It is also possible to set the desired number of decimal places for this p-value.

The default N of cases to be displayed in each cell has been changed. If a weight variable is specified for an analysis, the weighted N will be displayed, even if the box for unweighted N was checked on the option screen. However, the user can uncheck both boxes to have neither N displayed. Or the user can check both boxes to have both the weighted and the unweighted N displayed in each cell.

Charts are now available for MEANS.

• ## CHARTS

Charts have been added to the MEANS program output. The chart options are bar charts (the default) and line charts. The charts in MEANS work basically the same way as in the TABLES program.

The labels in charts have been improved. The labels for charts with many categories are less likely to be covered up. Also, the fonts used in charts can be changed. To change the font used for charts for a specific dataset, a particular font can be specified with the ’CHARTFONT=’ keyword in the datasets section of the HARC file. To change the default font for charts, a particular font can be specified in the servlet configuration file. The color palette(s) for the chart bars and lines can also be customized in the servlet configuration file. (On setting up the servlet configuration file, see the SDA 3.4 Archive Developer’s Guide for details.)

## LANGUAGE SUPPORT

New capabilities have been added to SDA in order to display properly a variety of languages. The new specifications can be found in the following places:

## QUICK TABLES

The ’Quick Tables’ program has been updated to use version 3.4 of the SDA TABLES program. Quick Tables was developed for ICPSR a few years ago and was not distributed routinely. However, Quick Tables is now included in the standard SDA distribution package.

CSM, UC Berkeley
January 25, 2010