olat_extract_html_results: Extracting OpenOLAT Results from HTML Results Files

In c403: Exam Tools for Department of Statistics (c403), Uni Innsbruck

Extracting OpenOLAT Results from HTML Results Files

Description

To get detailed information from the individual questions of a test, this function processes the HTML results files (test summaries) created by OpenOLAT, and (if required) combines this the data with the information provided by exams2openolat to create the link to the question source files (i.e., R/exams Rmd/Rnw questions). Is able to process results exported via Open Olat (zipfile) in both German or English for now.

Usage

olat_extract_html_results(rds, zipfile = NULL, verbose = TRUE)

Arguments

rds	NULL or character. If character, name of the RDS file procuded by exams2openolat when the test was generated used to extract original exam question file name and 'exname'. If NULL only the information from the zipfile is used.
zipfile	character, name/path to the ZIP file (exported via OpenOlat) oontaining the HTML results files (amongst other files needed).
verbose	logical, if TRUE some information is written to stdout.

Details

TODO(R): Note that this function is still under development and is missing a few features, see Sections 'Disclaimer' and 'Adding functionality'.

A brief summary of the individual steps:

1. Reading and flattening the rds file (source file name, and question name (exname)).

2. Unzips the zipfile, tries to guess the language of the content in the ZIP file, and checks if the content looks as expected.

3. Searching for the manifest (imsmanifest.xml) containing the OpenOLAT IDs; combine with the information from step (1) to create the link between source files and OpenOLAT IDs.

4. Identifying all HTML results files (HTML test summary) containing the required detailed information.

5. Parsing all HTML files, extracting user information and details about each individual question (OpenOLAT ID, Score, user Answer, correct Solution). This is combined with the information from step (2).

After that a few small modifications are made on the resulting data.frame before it is returned.

Value

Returns a (tibble) data frame of dimension ⁠N x K⁠ where N corresponds to the total number of questions. If 100 participants have taken a test with 20 questions each, this would result in N = 2000. The columns contain:

• Username: Participants user name (c-Kennung).

• Institution: Institution identifier (Matrikelnummer)

• Name: Full name of the participant.

• ID: OpenOLAT question identifier.

• Email: Participant email as shown in the HTML results file.

• Status: Status provided by OpenOLAT.

• Score: The participant's score for this question ("My Score" in the HTML file).

• Answer: The participant's answer, can be NA if empty.

• Solution: The correct solution.

• RawText: Raw question text, unformatted, can be used to search for keywords in the text.

• Section/Item: Section and item ID extracted from the exams olat identifier.

If an additional RDS file (argument rds) is provided, the following columns are added:

• file: Source file name (modified), path/name of the original R/exams file (Rmd/Rnw).

• exname: Name of the question (the exname meta-information from the R/exams question).

Besides Score (numeric), Section/Item (int) everything is of class character.

Disclaimer and missing features

TODO: This is a first implementation of this feature and there is a series of todo's and missing features. Incomplete list of known missing features:

• Depending on the (user specific) OpenOLAT language setting, the exported ZIP file comes in different languages. We 'auto guess' the language based on the content of the ZIP archive, tested for German and English.

• Originally implemented for string, num, and schoice questions. Not sure if it works for mchoice, definitively no support for cloze questions at the moment.

• Handling of multiple attempts (evaluate first, second, last attempt?).

• Support for multiple tests (if the "Results" folder contains multiple "test*" directories).

• Variations of questions (limited support; import works but post-processing might get difficult/impossible).

Adding functionality

TODO: The plan is to add additional functionality to produce output similar to what has been implemented already for the "nops workflow", to allow performing the same checks/analysis no matter where the results come from. See:

• https://git.uibk.ac.at/econstat/exams/demo/-/tree/main/gp-2024-02

Author(s)

Reto

Examples

## Not run: 

# 'Minimal' of the analysis performed in December 2024 (for which this
# function was originally written); few ggplot2 plots plus code to prepare
# the data matrix for estimating a Rasch model. Can't be run by end-users as
# we can't provide the data needed to run the example, this serves as a
# template for future developments/ideas.
library("c403")

rds     <- "GP06Dezember2024.rds"
zipfile <- "results_GP06Dezember2024.zip"
results <- olat_extract_html_results(rds, zipfile)


# ------------------------------------------------------------------
# Some plotting
# ------------------------------------------------------------------

library("ggplot2")
library("patchwork")

# Some demo plots; assumes that the students could only
# gain 0 points (incorrect) or 2 points (correct)
results$label <- with(results, paste(file, exname, sep = "\n"))
results$scoreLabel <- factor(results$Score > 0,
                             levels = c(FALSE, TRUE),
                             labels = c("Score == 0 (incorrect)",
                                        "Score > 0 (counted as correct)"))

g1 <- ggplot(results) + geom_bar(aes(y = label, fill = Status), stat = "count") +
          labs(title = "Stacked barplot of answered/unanswerd questions") +
          labs(subtitle = paste("Number of participants: ", length(unique(results$User)))) +
          theme_minimal()
g2 <- ggplot(results) + geom_bar(aes(y = label, fill = scoreLabel), stat = "count") +
          labs(title = "Stacked barplot of answered/unanswerd questions") +
          labs(subtitle = paste("Number of participants: ", length(unique(results$User)))) +
          theme_minimal()

plot(g1 + g2) # Patchwork plot

# Histogram of Score distribution; here assuming the threshold was 22 points
library("dplyr")
results %>% group_by(Name) %>% summarize(Score = sum(Score)) %>%
    ggplot() + geom_histogram(aes(x = Score), binwidth = 1) +
        geom_vline(xintercept = 22, col = "tomato", lwd = 2) + 
        labs(title = "Histogram of total Score") +
        labs(subtitle = paste("Number of participants: ", length(unique(results$User)))) +
        theme_minimal()

# ------------------------------------------------------------------
# Estimating Rasch model
# ------------------------------------------------------------------

library("tidyr")
library("psychotools")

# Prepare result to binary matrix for Rasch model
tmp <- transform(subset(results, select = c(Name, file, Score)), Score = as.integer(Score > 0)) |>
        pivot_wider(names_from = file, values_from = Score) |>
        as.data.frame()
tmp <- as.matrix(tmp[, !grepl("^Name$", names(tmp))])

# Estimate Rasch model
mr <- raschmodel(tmp)

# Default plots
hold <- par(no.readonly = TRUE)
par(mar = c(20, 12, 2.5, 1))
plot(mr, type = "profile")
par(hold)
plot(mr, type = "piplot")


## End(Not run)

c403 documentation built on March 6, 2026, 3:01 a.m.