gr0066_2. Speaking Stata: Multiple bar charts in table form. N. J. Cox. Stata Journal 17: 779; 16: 491–510.
New options frame() and frameopts() allow framing of bars, allowing so-called thermometer plots or charts. frame(
#
) specifies that all bars be framed with bars showing constant value #. Most commonly, # is the maximum possible value (say, fraction 1 or percent 100) or some other reference value. By default, framing bars have no fill color. frameopts() specifies options of twoway rbar to tune representation of the framing bars.
The help file has been updated to document these options and to include further references, both older and modern.
st0375_3: Estimating net survival using a life-table approach. E. Coviello, P. W. Dickman, K. Seppä, and A. Pokhrel. Stata Journal 19: 497; 18: 758–759; 15: 173–185.
In this new version, the stnet command allows users to specify two new options: unique creates intervals by cutting survival time at each observed failure; at(
numlist
) reduces the list of the results by showing net survival estimates only at specified survival times.
A minor bug has also been fixed.
st0383_1: Global search regression: A new automatic model-selection technique for cross-section, time-series, and panel-data regressions. P. Gluzmann and D. Panigo. Stata Journal 15: 325–349.
The new gsreg command is faster, robust, and more flexible. It fixes reported bugs, allows users to apply more estimators (from xtreg, logit, probit, and ivregress 2sls and ivregress liml), and works with weights. It also computes the out-ofsample root mean squared error with double precision. In addition, it saves intermediate results in Mata matrices to increase input and output operations speed (which strongly reduces global execution times).
st0449_1: Versatile tests for comparing survival curves based on weighted log-rank statistics. T. G. Karrison. Stata Journal 16: 678–690.
The verswlr command determines the significance level of the maximum of G
0,0, G
1,0, and G
0,1 weighted log-rank statistics or three other user-defined tests for the comparison of two survival curves. As Karrison (2016) notes, because weighted log-rank tests compare the hazard functions, an early emphasis test and a late emphasis test can both reject the null hypothesis in opposite directions. This can occur with crossing hazard functions regardless of whether that leads to crossing survival curves. Freidlen and Korn (2019) have constructed an example with crossing hazards in which the survival curve in the control group dominates the survival curve in the experimental arm, yet a test with increasing weights rejects the null hypothesis in favor of the experimental arm. To better alert users to such cases, the verswlr command now includes a statement in the output for each component test indicating which treatment arm is favored by that test. Users should therefore inspect the results of the individual component tests carefully, along with the survival curves, to identify such cases and interpret the results appropriately.
st0524_2: baselinetable: A command for creating oneand two-way tables of summary statistics. S. Donath. Stata Journal 18: 758–759; 327–344.
This update to the baselinetable command includes three new features:
fweights and aweights are now allowed.
String variables are now allowed for both categorical row variables and the optional by variable.
For categorical row variables, the totalrow option includes a row with column totals for the variable.
st0533_1. Univariate and multivariate outlier identification for skewed or heavy-tailed distributions. V. Verardi and C. Vermandele. Stata Journal 18: 517–532.
The gboxplot command has been superseded by a new command, robbox, that is available from the SSC archive (Jann et al. 2019). With robbox, many additional options are available, and standardization has been revised. The code also runs faster. Ben Jann is the main author of the new version of the code.
