Data sources
The data files Part One Chapters 1–4 and the R support pages use, with their columns and their reuse terms
This page is the site’s register of public data. It lists every data file Part One Chapters 1–4 and the R support pages use, says where each file comes from, what its columns mean, and how it may be reused. Two further records are kept as reference sets: they belong to no chapter but appear in worked examples. Chapters outside that range describe their own data in place: Part Two Chapter 2, for instance, gives the columns and the origin of the simulated file it offers for download, and the site’s licence gives its reuse terms. Private course data are never listed here as a download; they are opened locally in your own browser.
Find your chapter in the register below, then read that data set’s section. Each section follows the same order: what the file is, where to get it, the columns you will meet, the source and its attribution, the reuse terms, and anything that limits what the records can support. Every example on this site uses base R with the standard stats and graphics commands; no package is loaded or installed. Programming examples use short, meaningful names such as group, score and id; when records are changed, d0 holds the starting data and dat the working copy. Calculate at full precision and report statistical results rounded once to four decimal places, with trailing zeros dropped (3.1, not 3.1000); rounded inputs or different software rounding can change the last digits.
The register
| Chapter | Data set | File | Reuse |
|---|---|---|---|
| Part One, Chapter 1 | Student outcomes at a Portuguese polytechnic | realinho2022_students.csv, beside the chapter |
CC BY 4.0 |
| Part One, Chapter 2 | Secondary-school mathematics records | Not distributed; the values are inside the chapter | CC BY 4.0 |
| Part One, Chapter 3 | Special Practice | data/chapter_03_special_practice.csv |
CC0 1.0 |
| Part One, Chapter 3 | Student Performance extract | data/chapter_03_uci_groups.csv |
CC BY 4.0 |
| Part One, Chapter 4 | Background sound and reading comprehension | data/background_sound.csv |
CC0 1.0 |
| Part One, Chapter 4 | Feedback and practice quiz scores | data/feedback_practice.csv |
CC0 1.0 |
| Reference set | Reading instruction | Not distributed; from the carData package |
Package terms, GPL-2 or later |
| Reference set | Review timing | data/resource_timing.csv and its codebook |
See the section |
| Course data | Private course data | Not distributed; opened locally | Course terms |
Six CSV files can be downloaded from this page: five from the site’s data/ folder and the Chapter 1 file published beside its chapter. Each uses a comma as the delimiter and carries variable names in its first row.
Part One, Chapter 1: student outcomes
Chapter 1 follows a real administrative record of 4,424 students at a Portuguese polytechnic: 1,421 with recorded Dropout status, 794 still Enrolled, and 2,209 Graduate at the normal duration of their degree. The chapter uses it to ask how the chance of recorded dropout relates to admission grade and age at enrollment.
Where to get it. Download the student-outcome data (CSV), published beside Chapter 1. It holds three fields of the source release in the original case order, with no values changed.
| Field | Meaning and coding |
|---|---|
recorded_outcome |
Dropout, Enrolled or Graduate, the status recorded at the normal end of the degree. |
admission_grade |
The recorded admission grade, on the source’s own scale; the values here run from 95.0 to 190.0. |
age |
Age in whole years at enrollment; the values here run from 17 to 70. |
There are no missing values in the three fields.
Source and attribution. Realinho, V., Machado, J., Baptista, L., & Martins, M. V. (2022). Predicting student dropout and academic success. Data, 7(11), 146. https://doi.org/10.3390/data7110146. The release is archived as Realinho, V., Vieira Martins, M., Machado, J., & Baptista, L. (2021). Predict students’ dropout and academic success [Dataset]. UCI Machine Learning Repository. https://doi.org/10.24432/C5MC89. The release holds many more fields than the three published here.
What the records can support. Recorded Dropout can include a change of course or of institution; it need not mean leaving higher education altogether, and the three categories keep their distinct meanings. The chapter’s logistic regression is a simplified teaching analysis, not the source authors’ classification procedure. The release supplies neither the enrollment year nor dropout event dates, so a predictor’s availability at the time of prediction has to be argued rather than assumed.
Reuse. The source release and this three-field extract are licensed under Creative Commons Attribution 4.0. When redistributing or adapting the extract, credit Realinho and colleagues and the UCI Machine Learning Repository, keep the source and licence links, and identify the selection of fields as an adaptation together with any further change you make.
Part One, Chapter 2: secondary-school mathematics records
Chapter 2 describes 395 students of the mathematics file in Cortez and Silva’s student-performance study, drawn from school reports and questionnaires at two Portuguese secondary schools. The chapter asks how recorded absences and final mathematics grades vary together.
Where to get it. Not distributed as a separate file. The selected values are written into Chapter 2 itself, so the chapter renders without reading an external data file. The source files and their documentation are at the repository page linked below. A different three-column extract of the same source is distributed on this site for Chapter 3.
These are the columns the chapter works with.
| Field | Meaning and coding |
|---|---|
student_id |
Label U001 through U395, added for the chapter; it is not an original participant identifier. |
school |
GP or MS, the two secondary schools of the source records. |
absences |
Source field absences: the recorded number of school absences. |
grade_period_2 |
Source field G2: the second-period mathematics grade on the 0–20 scale. |
final_grade |
Source field G3: the final mathematics grade on the 0–20 scale. |
higher_education |
Source field higher: yes if the student reported an intention to pursue higher education, no otherwise. |
pass |
Derived from final_grade: Yes at a grade of at least 10 on the 0–20 scale, No below it. |
All 395 records are kept, in their source order, and the selected fields have no missing values.
Source and attribution. Cortez, P., & Silva, A. M. G. (2008). Using data mining to predict secondary school student performance. In A. Brito & J. Teixeira (Eds.), Proceedings of 5th Annual Future Business Technology Conference (pp. 5–12). EUROSIS. The data are archived as Cortez, P. (2008). Student Performance [Dataset]. UCI Machine Learning Repository. https://doi.org/10.24432/C5TG7T. The repository page supplies the source files and the variable documentation, including a separate Portuguese-language dataset that the chapter does not use.
What the records can support. The 395 students are the students the study observed. A summary calculated from them describes that collection directly; a statement about secondary students elsewhere needs an additional argument about how these students relate to that wider group. Higher-education intention was reported, not assigned.
Reuse. The source is licensed under Creative Commons Attribution 4.0. The chapter selects and renames variables, adds identifiers and a pass indicator, and keeps the selected values unchanged. Credit Cortez and Silva and the UCI Machine Learning Repository, keep the source and licence links, and identify those adaptations and any further change you make.
Part One, Chapter 3: Special Practice
This QMSBR record contains 120 fictional participants: 60 assigned to Special Practice and 60 to Usual Practice. The imagined study compares two practice formats with four sessions and equal total study time. A common assessment, scored from 0 to 40 points, takes place two weeks later. These are constructed teaching records, not observations from an actual experiment.
Where to get it. Download the Special Practice data (CSV).
| Field | Meaning and coding |
|---|---|
id |
Artificial row identifier, C3-001 through C3-120; it has no personal meaning. |
group |
Special Practice or Usual Practice, with 60 records in each condition. |
score |
Integer score on the common delayed assessment, in points on a 0–40 scale. |
All 120 records are retained, and the three fields contain no missing values. The comparison direction is Special Practice minus Usual Practice.
What the records can support. The assignment and follow-up details belong to the invented study description; the values alone do not establish a research design. The supplied rows define a fixed teaching record: calculate software results from those rows and keep their full precision during the analysis. Use these records for CSV tasks, and use the inputs supplied in an exercise for that exercise’s calculations.
Reuse. QMSBR dedicates this synthetic CSV to the public domain under CC0 1.0 Universal. The dedication applies to the CSV. Keep its description as synthetic when presenting or adapting it.
Part One, Chapter 3: Student Performance extract
This extract keeps all 395 records of the mathematics file in Paulo Cortez’s Student Performance dataset, reduced to four columns for the group comparison and the grade displays on the Chapter 3 page. The source records concern students in two Portuguese secondary schools. Higher-education intention was observed, not assigned. One row represents one recorded student. Chapter 2 of the book works with a wider selection of the same source.
Where to get it. Download the Student Performance extract (CSV).
| Field | Meaning and coding |
|---|---|
id |
Artificial identifier U001 through U395, added for this extract. It is not an original participant identifier. |
higher |
Source field higher. yes means the student reported an intention to pursue higher education; no means no such intention was reported. |
pass |
Derived from the source final grade G3. Yes means a grade of at least 10 on the 0–20 scale; No means a grade below 10. |
final_grade |
Source field G3, the final mathematics grade on the 0–20 scale, as recorded. |
All rows stay in their source order and the four fields have no missing values. There are 375 yes intention records and 20 no intention records. For a passing proportion within an intention group, use that group’s total as the denominator. The comparison direction is intention yes minus intention no, and the result is an observed association in this record.
Source and attribution. Cortez, P. (2008). Student Performance [Dataset]. UCI Machine Learning Repository. https://doi.org/10.24432/C5TG7T. The repository page supplies the source files and the variable documentation. QMSBR prepared the extract from the source mathematics file: it adds an artificial identifier, selects and renames two fields, keeps the final grade, and derives pass status from it.
What the records can support. The source documentation marks the dataset as having no missing values, but the source mathematics file contains an unresolved pattern: 38 records have a final grade of zero and zero absences, and 25 of those also have a positive second-period grade. The documentation does not establish a common meaning for all these zeros. The primary comparison keeps grades as recorded; omitting the zero-grade records is a diagnostic check that changes the analysed set, not an established correction. The extract keeps the final grade, so the zero-grade records can be identified, but it omits absences and second-period grades, so the pattern behind the zeros cannot be examined here. Read any supplied diagnostic result alongside this limitation.
Reuse. The source and this extract are licensed under Creative Commons Attribution 4.0. When redistributing or adapting the extract, credit Cortez and the UCI Machine Learning Repository, keep the source and licence links, and identify QMSBR’s adaptations and any further change you make. The licence does not imply endorsement by the source creators.
Part One, Chapter 4: background sound and reading comprehension
This QMSBR record contains 30 invented first-year university students. Each read the same passage for 20 minutes in an individual session, with one of three background-sound conditions playing, and then answered a 20-item reading-comprehension quiz scored from 0 to 20. The students were individually and randomly assigned, ten to each condition. The Chapter 4 page uses these records for its one-way ANOVA, its comparisons with Silence and Tukey’s pairwise intervals, and Welch’s one-way ANOVA.
Where to get it. Download the background-sound data (CSV).
| Field | Meaning and coding |
|---|---|
student |
Artificial identifier S01 through S30; it has no personal meaning. |
sound |
Silence, Instrumental music or Music with lyrics, in that order; Silence is the reference condition of the declared comparisons. |
score |
Integer comprehension score: the number of quiz items answered correctly, on a 0–20 scale. |
The rows run through Silence, Instrumental music and Music with lyrics in that order, ten rows for each condition. There are no missing values and no student appears twice. A CSV stores the labels as plain text, so R orders them alphabetically, placing Instrumental music first, unless the script declares the order. Every analysis sets levels = c("Silence", "Instrumental music", "Music with lyrics") with factor() before fitting a model.
What the records can support. The 30 scores are fixed authored values, not a sample drawn when the page is built. They place the three condition means two points apart with similar spreads, so every calculation can be reproduced by hand. The study, its random assignment and its scores are invented for teaching: they do not validate a Normal model or equal variances, and they are not evidence about any real effect of background sound on reading.
Reuse. QMSBR dedicates this synthetic CSV to the public domain under CC0 1.0 Universal. Keep its description as invented teaching data when presenting or adapting it.
Part One, Chapter 4: feedback and practice quiz scores
This QMSBR record contains 24 invented Grade 7 students, each measured once on a fractions quiz scored from 0 to 20. The imagined study crosses three feedback conditions with two practice schedules, giving six cells of four different students. The Chapter 4 page uses all 24 rows for its factorial model, its interaction plot and the declared comparison; its one-way example uses the background-sound record.
Where to get it. Download the feedback and practice data (CSV).
| Field | Meaning and coding |
|---|---|
student |
Sequential identifier 1–24; it has no personal meaning. |
feedback |
Rereading, Answer feedback or Explanatory feedback, in that teaching order. |
practice |
Massed or Spaced, in that teaching order. |
score |
Integer quiz score in points on a 0–20 scale. |
There are no missing values and no student appears twice. A CSV stores the labels as plain text, so R orders them alphabetically unless the script declares the order. Every analysis sets levels = c("Rereading", "Answer feedback", "Explanatory feedback") and levels = c("Massed", "Spaced") with factor() before fitting a model.
What the records can support. The 24 scores are fixed authored values, not a sample drawn when the page is built, and the repeated residual pattern is deliberate so that the arithmetic stays visible. They are teaching records: they do not validate a Normal model, equal variances, or any real effect of feedback or practice scheduling.
Reuse. QMSBR dedicates this synthetic CSV to the public domain under CC0 1.0 Universal. Keep its description as invented teaching data when presenting or adapting it.
Reference set: reading instruction
This reference example uses Baumann from carData: Companion to Applied Regression Data Sets, by John Fox, Sanford Weisberg, and Brad Price. The package documentation for Baumann describes 66 records with a group field, two pretests, and three post-tests. It traces the teaching dataset to Moore and McCabe’s Introduction to the Practice of Statistics (2nd ed., 1993, pp. 794–795).
Where to get it. Not distributed here, and no page on this site loads the package. Install carData in desktop R, then load the dataset explicitly and prepare a working copy with the example’s group labels:
data("Baumann", package = "carData")
d0 <- Baumann
dat <- d0
dat$score <- dat$post.test.1
dat$group <- factor(
dat$group,
levels = c("Basal", "DRTA", "Strat"),
labels = c("DRA", "DRTA", "Think-aloud")
)This code keeps the package record in d0, copies it to dat, gives the selected outcome the short name score, and changes the working group labels. It retains all 66 observed records and their scores. The outcome used here is post.test.1, the selected error-detection score in points on a 0–16 scale. The original package object remains available as Baumann.
| Package label | Example label | Assigned children | Observed outcomes |
|---|---|---|---|
Basal |
DRA: directed reading activity | 22 | 22 |
DRTA |
DRTA: directed reading-thinking activity | 23 | 22 |
Strat |
Think-aloud instruction | 23 | 22 |
Source and attribution. The study description and assignment counts come from Baumann, Seifert-Kessell, and Jones (1992), Effect of Think-Aloud Instruction on Elementary Students’ Comprehension Monitoring Abilities, particularly the Method section on printed page 147. The study assigned 68 fourth-grade children in one rural Midwestern school. One child in each innovative condition had an unavailable outcome because of illness or transfer. The package therefore contains 66 observed outcomes, with 22 per group; it provides no scores for the two unavailable outcomes.
What the records can support. The reference example selects one outcome and uses an unadjusted one-way model with a common within-group variance. The source study used several outcomes and pretest-adjusted analyses. Its integrated success criterion is invented for an imagined future replication: both DRA-referenced contrasts must have lower endpoints above zero in a two-sided 95% Bonferroni family. Applying that criterion to these historical records is a post hoc teaching analysis, not a prespecified decision attributed to the original investigators. Linking complete-outcome calculations to effects for all assigned children also requires a justification for the unavailable outcomes; a completion check alone does not establish that link.
Reuse. This page distributes no Baumann CSV and no copy of the package. The package keeps its upstream GPL version 2 or later terms, as stated on its CRAN page. Anyone redistributing the package or an extract derived from it should keep the applicable upstream terms and notices. The research article is cited, not redistributed, and its publication terms are separate from the package’s terms.
Reference set: review timing
This 36-case file keeps every score, factor code and case order of an archived two-factor teaching file, whose columns were Score, A and B. A sequential identifier was added and the two factors were given the labels below. The Grade 7 fractions lesson, the resource descriptions, the review times, the quiz interpretation and the random assignment are invented for Chapter 4. The archived scores themselves were not newly simulated, and the source file documents neither a real educational study nor a randomization.
Where to get it. Download the review-timing data (CSV) · Full codebook
| Field | Meaning and coding |
|---|---|
student |
Added sequential case identifier, 1–36. |
resource |
Original A: 1 = Textbook, 2 = Tutorial, 3 = Video. |
timing |
Original B: 1 = Same day, 2 = Two days, 3 = One week. |
score |
Original Score, unchanged. Values range from 7 to 20; the invented setting reads them as numbers correct on a 20-item retention quiz. |
There are four complete cases in each of the nine cells and no missing values. Scores sum to 407. In resource order Textbook, Tutorial, Video and timing order Same day, Two days, One week, the cell means are (8.50, 9.00, 8.25), (15.75, 14.25, 18.25) and (9.75, 9.75, 8.25).
What the records can support. A full resource-by-timing model fits these records under independent-student, common-variance working assumptions, and Normal errors give the exact classical F and t references. These archived scores are kept as a reference set; the Chapter 4 page uses the background-sound and feedback and practice scores instead.
Reuse. Keep the archived-file attribution and identify the educational setting as invented when adapting this example. The CC0 dedication for the Special Practice CSV does not apply to these archived scores. This page assigns no new licence to the archived file; QMSBR’s own descriptions and code keep the terms in LICENSE.
Private course data
The HSB exercises use CSVs distributed through a course. These pages give the import and analysis instructions, including the minimal field names and historical category codes each activity needs, and they distribute no HSB records and no HSB results. Open your own file locally in the workspace; it stays in your browser and is not uploaded to this website. The complete dictionary and the provenance qualifications accompany the course’s own data reference.