README.md

In dacarras/r4sda: Collection of functions to produce descriptives analysis of secondary data and large scale assessments studies

r4sda

• The goal of r4sda is to make it easier to work with data from large scale assessment studies.
• Is a collection of simple wrapper functions, mainly, generated to solve common tasks when working with data from large scale assessment, secondary data from large national records, or other examples of data with nested observations
• It generates tables, variables, weights, and other objects to aid data analysis
• It relies on libraries such as: dplyr, stringr, purrr among others

Installation

You can install, in R, the development version of r4sda from this github repository with:

devtools::install_github("dacarras/r4sda")

Development

Pendings

• add sample data to show functions examples
• add a viggnete
• add merge function

Done

• add functions for weights in mixed models, methods a and method b, also called normalized and effective sample size weights
• add mean score within observations
• add check cluster id function

List of functions

Descriptives

• wide_resp() generates a table of items as rows, and response value as columns, and displays the percentage of responses per item.
• wide_var() generates a table of response as rows, and variablesas columns, and displays the percentage of responses per item.
• stack_resp() generates a table of items as rows, responses as attributes of items, and the percentage of each response category.
• get_desc() it produces a table with descriptives, including: percentage of missing, complete observations, n, means, sd, minimum, maximum, skewness, kurtosis, and histogram of variables.

Aggregated scores

• c_mean() estimate cluster means to aid cluster mean centering in mixed models.
• c_wmean() estimate cluster means to aid cluster mean centering in mixed models, including weights within clusters.
• c_sum() estimate cluster sums to aid cluster variables generation for mixed models.
• c_sd() estimate cluster standard deviations of a variable.

Manifest Scores

• reverse() it generates a reverse score for any given numeric vector (it removes the labels if the variable is labelled).
• z_score() it standardize variables returning these as z scores.
• mean_score() it create mean score of variable (i.e. row wise means).
• sum_score() it create sum score of variables (i.e. row wise sum).

LSA Weights

• lsa_weights() add normalized and effective sample weights to the provided data frame.
• senate_weights() add senate weights scaled up a to a number (e.g. 500, 1000 or else).
• jkr_iccs() add jackknifes replicate weights to ICCS 2009 study data frame.
• svy_freq() estimates proportions for each category from categorical variable, from a survey object.

Meta-data

• variable_label() it gets variable labels from a variable, from labelled vector.
• value_label() it generate a table from the value labels from a labelled vector.
• variables_table() generates a table of a data frame, including variable names, variable types, sample values, and variable labels.
• remove_labels() it remove labels from a data frame. It aids the use of data frame for other packages and software that needs plain data for their use.

Mixed Models

• get_icc() it estimates the Intra class correlation of an MLM model from lme4.
• check_cluster_id() it tests if the cluster id is unique across the data frame, or if these repeats between addtional cluster factors. For example, it checks if schools id repeats between country observations, or if observations id are unique between schools.
• caterpillar_plot() it extracts the realizations of a random intercept model, generated by lme4. The output is a plot, from ggplot2, thus, the user can further specify theme options, axis length, among other customizations. It was develop to visually inspect random intercept spreadings, for unconditioned and conditioned models.
• caterpillar_mean_plot() it extracts the realizations of a random intercept model, generated by lme4. The output is a plot, from ggplot2, thus, the user can further specify theme options, axis length, among other customizations. It was develop to visually inspect random intercept spreadings, for unconditioned and conditioned models. This version adds the grand of the model to the random effect, depicting the latent mean in return.
• center_variables() generate centered variables for mixed models for a covariate. It provides the centered within cluster (cwc) and the cluster mean center to the grand mean components. This two components are needed to fit disaggregated models. Moreover, the cwc is use to fit random slope models.
• […]

Misc

• local_path() aids the generation of relative working folders. It assumes your syntax and data folders are on the same logical level of folder structure, and aids getting the parent directory. As such:
• local_path(getwd()) retrieves the parent directory of workint directory
• local_path('/00_data/') retrieves the absolute location of the ‘/00_data/’ folder, if this folder is below the parent directory of the working directory, where the syntax is located.
• This function allows to run ’*.rmd’ files between Mac and Windows machines using Rstudio, or R-Box, and source the entire code using a relative folder structure.
• An example of such a structure is:

/00_data/
/01_syntax/
/02_tables/

• text_to_table() helps to take unstructured text that includes the variables of MPLUS data, and create a table with the list of included variables.

• decimal() format number with decimal places, into strings with a given number of decimal places.

• get_lrt_scf() Likelihood Ratio Test for MPLUS fitted model with the MLR estimator (see https://www.statmodel.com/chidiff.shtml.)

Example

This is a basic example which shows you how to solve a common problem:

# create an items data frame
# items_data <- dplyr::select(data_frame, LS2T01:LS2T16)

# request a wide response data table with wide_resp() function
r4sda::wide_resp(items_data)

   variable    `1`    `2`    `3`    `4`    `NA`
   <chr>     <dbl>  <dbl>  <dbl>  <dbl>   <dbl>
 1 LS2T01   0.105  0.308  0.0917 0.486  0.00961
 2 LS2T02   0.697  0.103  0.150  0.0389 0.0104 
 3 LS2T03   0.133  0.0316 0.758  0.0708 0.00664
 4 LS2T04   0.100  0.115  0.0423 0.734  0.00841
 5 LS2T05   0.567  0.0802 0.200  0.147  0.00641
 6 LS2T06   0.276  0.219  0.232  0.250  0.0234 
 7 LS2T07   0.0951 0.213  0.494  0.190  0.00871
 8 LS2T08   0.138  0.237  0.584  0.0317 0.00945
 9 LS2T09   0.191  0.426  0.104  0.270  0.00808
10 LS2T10   0.382  0.187  0.202  0.209  0.0208 
11 LS2T11   0.447  0.271  0.107  0.162  0.0125 
12 LS2T12   0.274  0.242  0.159  0.309  0.0150 
13 LS2T13   0.247  0.417  0.149  0.170  0.0166 
14 LS2T14   0.600  0.159  0.0940 0.129  0.0187 
15 LS2T15   0.0956 0.134  0.0665 0.683  0.0210 
16 LS2T16   0.0801 0.593  0.0948 0.208  0.0240 

Because this is a table, it can be render as kable table:

knitr::kable(wide_resp(items_data), digits = 2)

| variable | 1 | 2 | 3 | 4 | NA | |:---------|-----:|-----:|-----:|-----:|-----:| | LS2T01 | 0.11 | 0.31 | 0.09 | 0.49 | 0.01 | | LS2T02 | 0.70 | 0.10 | 0.15 | 0.04 | 0.01 | | LS2T03 | 0.13 | 0.03 | 0.76 | 0.07 | 0.01 | | LS2T04 | 0.10 | 0.12 | 0.04 | 0.73 | 0.01 | | LS2T05 | 0.57 | 0.08 | 0.20 | 0.15 | 0.01 | | LS2T06 | 0.28 | 0.22 | 0.23 | 0.25 | 0.02 | | LS2T07 | 0.10 | 0.21 | 0.49 | 0.19 | 0.01 | | LS2T08 | 0.14 | 0.24 | 0.58 | 0.03 | 0.01 | | LS2T09 | 0.19 | 0.43 | 0.10 | 0.27 | 0.01 | | LS2T10 | 0.38 | 0.19 | 0.20 | 0.21 | 0.02 | | LS2T11 | 0.45 | 0.27 | 0.11 | 0.16 | 0.01 | | LS2T12 | 0.27 | 0.24 | 0.16 | 0.31 | 0.02 | | LS2T13 | 0.25 | 0.42 | 0.15 | 0.17 | 0.02 | | LS2T14 | 0.60 | 0.16 | 0.09 | 0.13 | 0.02 | | LS2T15 | 0.10 | 0.13 | 0.07 | 0.68 | 0.02 | | LS2T16 | 0.08 | 0.59 | 0.09 | 0.21 | 0.02 |

dacarras/r4sda documentation built on Nov. 9, 2023, 10:17 a.m.