Added links to the IDC website and user guide.
For an attempted census, when the app creates replicate weights, it uses the random-groups jackknife method. This should drastically increase speed.
For the analysis type comparing estimates to external benchmark data, user-entered percentages must sum to 100.
The progress bar now appears for every analysis type in the app.
Informative error messages are shown for most analysis types when the user supplies a dataset with only respondents in it.
The minimum required version of the ‘nrba’ package is now 0.2.0.
The Setup module of the app now checks for bad inputs and displays informative error messages.
Whenever an error message appears for a specific analysis type, a modal dialog appears with suggestions for how to resolve common errors for that analysis type.
The Excel output file is formatted using Excel table styles. Important columns in the output table now have bolded entries to help guide users in interpreting the content.
The analysis types based on regression analysis now clearly indicate in the footnotes which category is used as the reference level for each categorical predictor variable.
In the weighting analysis type, the output table footnote clearly documents which variables were used to conduct the weighting adjustment.
Now using the survey package options options(survey.lonely.psu = 'adjust') and options(survey.adjust.domain.lonely = TRUE), so that domains with only a single PSU will have nonzero variance estimates.
If the user specifies strata population sizes but doesn’t specify any weights, then the weights are assumed to all be equal (i.e., equal to 1).
Minor updates:
Tooltips now appear throughout the app, even before loading a data file.
Aside from p-values, the default number of decimals for statistics is now 3 instead of 5.
Minor layout improvements for some analysis types.
Minor cosmetic improvements and typo fixes.
Simplified user interface for analyzing an attempted census.
Simplified user interface for the analysis comparing respondents to the full sample.
Removed ability to specify replicate weights to use for analysis.
Improved formatting so that percentages in output tables are consistently formatted as such.
Bug fixes:
Secondary Sampling Units (SSUs) are now correctly incorporated into variance estimation, if specified.
For regression analysis type, missing values could sometimes be treated as a valid non-missing category when the user is asked which category of the outcome variable they want to predict.