This update focuses mainly on bugfixes and improvements in the GUI and its workflow, but also introduces some new features.
All analysis functions, as well lsa.data.diag,
lsa.recode.vars and lsa.vars.dict crash with
an error message
Error in exists(all.vars(match.call())) : first argument has length > 1
when any of the arguments specifying analysis variables point to objects
which contain character vectors containing them. Thanks to Rodolfo
Ilizaliturri.
lsa.prctls - The percentiles on the x-axis are not
properly added for every percentile computed. Thanks to Cecilia
Bjorkhammer.
print.lsa.data - The custom print method for
lsa.data blocks some common data.table
operations.
lsa.corr - The function crashes when one of the
countries has all missing values for one or more variables passed to the
bckg.corr.vars. A warning when this occurs was
added.
GUI with lsa.convert.data - When data files only
from one country are available in the source folder, the quote at the
end of the path passed to inp.folder in the syntax is not
closed and the syntax cannot be executed properly. Thanks to Gasper
Cankar.
GUI with lsa.merge.data - When data files only from
one country are available in the source folder, the bracket at the end
of the list passed to file.types is not closed and the
syntax cannot be executed properly. Thanks to multiple teachers during
series of workshops delivered in Slovenia.
GUI with lsa.recode.vars - When new variable names
are defined, the table for the new variable labels still displays the
old variable names, as shown below. Thanks to Cecilia
Bjorkhammer.
GUI with lsa.data.diag - When the working data set
is changed, series of errors are dropped in the R console
(missing value where TRUE/FALSE is needed).
GUI with lsa.pcts.means - When the working data set
is changed, an error is dropped in the R console
(Error in paste0: object 'Variables' not found) and shown
in the interface.
GUI with lsa.prctls - When the working data set is
changed, an error is dropped in the R console
(Error in if: argument is of length zero) and shown in the
interface.
GUI with lsa.bench - When a data set is loaded or
the working data set is changed, an error is dropped in the R console
(Error in paste0: object 'Variables' not found).
GUI with lsa.crosstabs - When a data set is loaded
or the working data set is changed, an error is dropped in the R console
(Error in if: argument is of length zero).
GUI with lsa.corr - When the working data set is
changed, an error is dropped in the R console
(Error in if: argument is of length zero) and shown in the
GUI.
GUI with lsa.lin.reg - When the working data set is
changed, errors are dropped in the R console
(Error in if: argument is of length zero and
Warning: object 'Variables' not found) and shown in the
GUI.
GUI with lsa.bin.log.reg - When the working data set
is changed, errors are dropped in the R console
(Error in if: argument is of length zero and
Warning: object 'Variables' not found).
The lsa.crosstabs function now has the possibility
to produce heatmap graphs (optional). The graphs are included in a
separate sheet in the MS Excel output file if
save.output = TRUE. If save.output = FALSE,
the graphs are added to the list output object in memory and can be
printed in R’s graphic device.
All GUI tabs for data preparation and analysis functions received “Save syntax” and “Copy syntax” buttons, as suggested by Erika Majoros.
Further work on graphics (lsa.crosstabs,
lsa.bench, lsa.pcts.means,
lsa.prctls):
All functions producing graphical representation of the results
now have controlled dpi depending on the number of
split.vars to better fit the plots and their labels in
space;
The lsa.prctls function now adds facets when the
number of split.vars is greater than two to fix the issue
of dot, line and error bar positioning and overlapping on the plot and
improve readability.
Multiple fixes in GUI for better workflow.
In GUI with lsa.recode.vars the “Old/New variable
names” table was merged with the “New variable labels” table. This
change was provoked by fixing the bug that the “New variable labels”
table was still showing the old variable names, as discovered by Cecilia
Björkhammer. Now the workflow is more intuitive.
Improved documentation.
lsa.pcts.means - when many split variables are added,
in some countries some of the rows in the estimates are repeated
multiple times by the categories of the split variables.The lsa.pcts.means, lsa.bench and
lsa.prctls functions now have the possibility to produce
graphs (optional). The graphs are included in a separate sheet in the MS
Excel output file if save.output = TRUE. If
save.output = FALSE, the graphs are added to the list
output object in memory and can be printed in R’s graphic
device.
The MS Excel output files from all analysis functions received an additional sheet with warnings (if any) related with the computations. So far these warnings were just printed on screen.
Internal reorganization of the code producing warnings related with the computations and the way they are issued.
Improved documentation.
Various code changes following the update of R to version 4.2.0.
lsa.pcts.means does not show the syntax when
only splitting variables are chosen for the analysis and the GUI freezes
when the analysis is ran.lsa.pcts.means received a new argument,
central.tendency, which allows users to compute either the
arithmetic mean (default and available so far), median (new) or mode
(new) for continuous variables.
New function, lsa.select.countries.PISA, a utility
function that allows the user to select countries of interest from a
converted PISA data file (or PISA object residing in memory) and remove
the rest of the countries’ data. This is useful when the user does not
want to analyze all countries data available in an original a PISA data
file.
All analysis functions received a new argument,
save.output. If TRUE (default), the output is
written into MS Excel file, as it was so far. If FALSE, the
output (a list of all different estimates) is printed on screen or can
be assigned to an object. The argument is available in command-line use,
but not in the GUI.
Multiple fixes in GUI for better workflow.
Improved documentation.
All analysis functions. When “IDCNTRY” is added explicitly to the
list of split.vars and include.missing = TRUE,
the function crashes with the following error message. Thanks to Laura
Ringiene.
lsa.prctls function. When computations involve PVs,
the estimates of specific percentile appear in a wrong column, e.g. the
25th percentile columns contain the estimates for the 5th percentile
columns and the other way around. Thanks to Laura Ringiene.
lsa.data.diag function. When the tables are exported
to Excel, the “Index” sheet appears as the first one, but the file opens
with the sheet with statistics for the first variable on the
front.
GUI with lsa.data.diag and all analysis functions.
When using the GUI and an output file with the same name as the
specified is already opened, the GUI crashes when the respective
function tries to write the new output file.
lsa.bin.log.reg function. The coefficients output is
not sorted properly by the names of the independent variables and the
desired order.
GUI with lsa.corr function. When the loaded data
file in “Correlations” analysis contains no PVs, the radio buttons for
choosing between Pearson and Spearman correlation are not
shown.
GUI with all functions. If any error occurs during a function execution (i.e. after pressing the “Execute syntax” button), the GUI crashes.
New analysis function, lsa.crosstabs. It computes a
two-way table (crosstabulations) and Rao-Scott first- and second-order
design corrected chi-square statistics. The Rao-Scott adjustment is
needed because of the clustered design of the large-scale assessments
and surveys data.
Added the possibility to include two-way interaction terms in
lsa.lin.reg and lsa.bin.log.reg. The
interactions can be between two categorical variables, categorical and
continuous, continuous and continuous, categorical with PVs, continuous
with PVs, and PVs with PVs.
Added a new printing method for lsa.data objects in
memory. It prints the study name, cycle, respondent type(s), total
number of countries, key, if the data has user-defined missing values or
not, and a snippet of the data.
Added support for the recently released IEA eTIMSS 2019 Problem Solving and Inquiry (PSI) tasks’ data.
The RALSA logos in the GUI and GUI start-up screen was replaced with higher-resolution images.
Various visual improvements of the GUI.
Updated and improved documentation.
This is a maintenance version fixing some bugs and removing package dependency.
lsa.pcts.means function crashes with an error
message when no split variables are added and both background variables
and PVs are added to compute means for. Thanks to an anonymous RALSA
user.
lsa.bin.log.reg function crashes when the variable
name passed to bin.dep.var contains numeric
characters.
lsa.lin.reg function crashes when the variable name
passed to bckg.dep.var contains numeric
characters.
When used in the GUI, the lsa.recode.vars assigns
the new value labels to the recoded values in incorrect order. Thanks to
multiple workshop participants at the ECER 2021 pre-conference
training.
The gdata package is no longer needed by
RALSA.
Start up and warning messages when dependencies are loaded during GUI start are not displayed anymore.
This is a maintenance version following the update of base R to v4.1.0 where some functions’ behavior has changed and cause crashes in the analysis functions.
All analysis functions. Some functions in base R were updated and
their behavior has changed causing crashes in the analysis functions
with first argument has length \> 1 error
messages.
lsa.vars.dict function. When there are just two
levels for a factor variable, the second level in ?Variable levels? ends
without a single quote.
Occasionally, the lsa.data.diag freezes R with a
question mark in the console and eventually crashes the R session when
split.vars are provided.
Various optimizations in the lsa.corr,
lsa.lin.reg and lsa.bin.log.reg
functions.
The lsa.data.diag function now has user interruption
handling with a message.
All analysis functions. Any analysis function crashes when TALIS 3S data is used and the weight variable is specified explicitly.
All analysis functions. When ICCS teacher and school data are merged, the default weight, jakknifing zone and jakknifing replication indicator are not automatically detected and the function crashes, while specifying the weight manually passes.
lsa.prctls function. The order of the columns for
the percentiles, their SE, SVR and MVR is scrambled.
All data preparation and analysis functions. When using
data.object to provide data and the object is in quotes,
the function crashes with uninterpretable error message. Added custom
handle and error message.
lsa.lin.reg function. The function crashes when
using specific variable combinations.
GUI. The filtering of the tables with country ISOs, country names, variable names and labels is very slow and oftentimes unresponsive.
GUI with lsa.recode.vars. Occasionally, when the
lsa.recode.vars function is executed by pressing the
“Execute syntax” button, the interface crashes. This is because there is
a factor level for which no actual values exist in the data. Now handled
with custom error message.
lsa.merge.data function. The school weight was
dropped when merging school and teacher data in TALIS which prevented
the use of the merged file in multilevel models. Thanks to Jelena
Veletic.
lsa.data.diag, a quick automated production of weighted or
unweighted frequency tables for categorical variables and descriptive
statistics for continuous variables. These tables are for data
diagnostic purposes prior to analysis and are not for reporting results
from large-scale assessments.After updating R to v4.0.5 and the packages RALSA depends on, the warning messages in the GUI console were not displayed, although displayed in RStudio console. The warning messages are now back in the GUI console as well.
Showing and hiding elements in the GUI when using the benchmarks analysis type was improved.
The version number was incremented to 1.0 since the package has already achieved stability and maturity.
lsa.convert.data function. The function crashes when
the different cycles of the study change the case of the file names and
their extensions. Thanks to Yuan-Ling (Linda) Liaw.
lsa.recode.vars function. If the recoding
instruction ends with semicolon, the function crashes.
lsa.bin.log.reg function. The function crashes if
some specific patterns are found in the binary dependent variable
names.
lsa.pcts.means function. Some missing estimates in
columns Mean, Mean_SE, Mean_SVR, Mean_MVR, SD_SE and SD_MVR when the
groups are too small.
All analysis functions crash using SITES 2006 data when only mathematics teacher or only science teacher data are used.
lsa.pcts.means function. Some estimates using CivED
data were incorrect.
lsa.pcts.means function. Missing estimates in some
columns for specific combination of splitting variables.
The ISO argument in lsa.convert.data is
now case-insensitive.
The file.types and ISO arguments in
lsa.merge.data are now case insensitive.
The error messages in lsa.recode.vars are
improved.
lsa.convert.data function. Unrecognized characters
in factor levels are now fixed. Such were, for example the levels of the
number of books variable in PIRLS 2016 and other cycles which displayed
unrecognized characters instead of instead of “-”.
GUI. When categorical variables are added in the
list of Independent background categorical variables in
linear and logistic regression, the number of categories
(N cat. column) and the drop-down menu for the reference
categories (Ref. cat. column) include the missing values as
well.
lsa.convert.data function. For some variables
categories have the same labels as the missing ones in other variables
and are improperly converted as missing.
When loading or switching to a tab in the GUI, it is
scrolled to the position where the previous tab was scrolled
to.
TIMSS 2019 is now fully supported.
PISA for Development is now supported, as requested by David Joseph Rutkowski.
Various improvements for the GUI elements
location.
Improved documentation.
Links to the documentation for each functionality
RALSA supports were added to the Help section
of the GUI.
Improved messages, warnings and error messages.
The first version of the R Analyzer for Large-Scale Assessments
(RALSA) is released. RALSA targets both the
experienced R users, as well as those less technical skills. Thus, along
with the traditional command-line R interface, a Graphical User
Interface is featured.
Note that this is a “first release” version, so some bugs are expected.
RALSA is is used for preparation and analysis of data
from large-scale assessments and surveys which use complex sampling and
assessment design. Currently, RALSA supports a number of
studies with different design and a number of analysis types (see
below). Both of these will increase in future.
RALSA is a free and open source software licensed under
GPL v2.0.
Currently, RALSA supports the following
functionality:
Prepare data for analysis
Convert data (SPSS, or text in case of PISA prior 2015)
Merge study data files from different countries and/or respondents
View variable properties (name, class, variable label, response categories/unique values, user-defined missing values)
Recode variables
Perform analyses (more analysis types will be added in future)
Percentages of respondents in certain groups and averages on variables of interest, per group
Percentiles of variables within groups of respondents
Percentages of respondents reaching or surpassing benchmarks of achievement
Correlations (Pearson or Spearman)
Linear regression
Binary logistic regression
All data preparation and analysis functions automatically recognize the study design and apply the appropriate techniques to handle the complex sampling assessment design issues, while giving freedom to tweak the analysis (e.g. change the default weight, apply the “shortcut” method in TIMSS and PIRLS, and so on).
Currently, RALSA can work with data from all cycles of
the following studies (more will be added in future):
CivED
ICCS
ICILS
RLII
PIRLS (including PIRLS Literacy and ePIRLS)
TIMSS (including TIMSS Numeracy, eTIMSS will be added with the upcoming release of TIMSS 2019)
TiPi (TIMSS and PIRLS joint study)
TIMSS Advanced
SITES
TEDS-M
PISA
TALIS
TALIS Starting Strong Survey (a.k.a. TALIS 3S)