Using Small-Group Sound Games to Strengthen Early Literacy in Luganda

Outcome summaries show larger short-term gains in the intervention arm than in the comparison arm for both blending and segmentation.i. Data generation procedures, processing methods, devices, and toolsData generation proceduresThe workbook documents a multi-source data collection process.First, schools were approached and recruited. The Schools sheet records 10 approached schools in the Central Zone, of which 8 agreed and 2 declined because they could not add a supplemental activity to an already crowded timetable.

The 8 participating P2 classrooms were randomized at classroom level: 4 to the intervention arm and 4 to the comparison arm.Second, learner recruitment was recorded in the Learner_Roster sheet. Each learner record includes eligibility, caregiver consent, child assent, whether the assent script was administered in Luganda, baseline completion, post-test completion, missing-post-test reason, gender, age in years, and assessor IDs.Third, Luganda phonological assessment data were generated using a Learner Assessment Instrument.

The instrument has two scored subtasks:SubtaskNumber of scored itemsItem scoreTotal scoreBlending120/10–12Segmentation120/10–12The Assessment_Items sheet lists the item stimuli and target words. The Assessment_Raw Data sheet contains one row per learner assessment event, with separate baseline and post-test records. Item-level columns are coded as 1 = correct and 0 = incorrect/no response.

Blending and segmentation totals are calculated as sums of the 12 relevant item columns. The workbook notes that Luganda digraphs such as ny were treated as one sound.Fourth, fieldwork administration was documented in the Fieldwork_Log sheet. This records assessment phase, assessor ID, administration language, whether the assessment was conducted in an individual quiet space, masking status, whether the child referred to the games, envelope batch ID, whether the assessment sheet was sealed, and whether it was handed to the project coordinator.Fifth, implementation data were generated through teacher/session logs in the Session_Log sheet.

The intervention consisted of 8 weeks of implementation, with 3 planned small-group sessions per week in each of the 4 intervention classrooms, giving 96 planned sessions. Each session row records week, planned session number, game type, session status, dose counted, planned minutes, actual minutes, target group size, materials used, session timing relative to the regular Luganda block, reason for missed/shortened sessions, and teacher notes.Sixth, coach/observer fidelity data were collected in the Fidelity_Raw Data sheet.

Each row is an observation of a small-group sound-game session. The sheet records six binary fidelity indicators, target group size, game type, child engagement rating, and observer notes. The six fidelity indicators are:Teacher modelled the target skill clearly.Teacher used familiar Luganda words.Teacher followed the “I do, We do, You do” sequence.Teacher gave individual turns rather than only chorus answers.Teacher corrected errors explicitly and briefly.Teacher kept the session within the planned 20 minutes.Seventh, qualitative acceptability and implementation data were generated through teacher interviews.

The Teacher_Interview_Codes sheet contains coded interview excerpts or summaries, theme codes, valence, coded acceptability, coded ease of delivery, coded barrier, and acceptability rating where applicable.Devices, instruments, and toolsThe workbook identifies the following study instruments and implementation tools:CategoryTool or instrumentLearner outcome measurementLuganda Learner Assessment Instrument for blending and segmentationFieldwork quality controlAssessor guidance, sealed envelopes, daily handover to project coordinatorIntervention implementationSmall-group Luganda sound gamesClassroom materialsWord cards, oral prompts, clapping routines, bottle tops, chalk sound boxes, chalk board groupsFidelity measurementObservation and Fidelity ChecklistEngagement measurementChild engagement rating scale, 1 = very low to 5 = very highQualitative dataTeacher semi-structured interview guideData storage/analysisExcel workbook with raw sheets, formula-based analysis sheets, and overview sheetThe workbook does not specify use of tablets, phones, audio recorders, statistical software, or other electronic data-capture devices.

The fieldwork records point instead to paper-based assessment sheets handled through sealed envelopes and later compiled in the workbook.Data processing methods and stepsThe workbook uses a transparent, mostly formula-based processing structure.Assessment scoring Blending item scores B1–B12 are summed to produce a blending total from 0 to 12. Segmentation item scores S1–S12 are summed to produce a segmentation total from 0 to 12.Feasibility indicators The Analysis_Feasibility sheet calculates recruitment, consent/assent, retention, dose delivered, fidelity, teacher acceptability, and child engagement using numerator/denominator formulas and pre-specified feasibility criteria.Dose calculation In the session log, both “Delivered” and “Shortened” sessions count toward dose delivered. “Missed” sessions have dose counted as 0.Fidelity calculation Six binary fidelity indicators are summed across 64 observed sessions.

The overall fidelity denominator is 64 observations × 6 indicators = 384 indicator opportunities. The workbook reports 327 successful fidelity indicators out of 384, or 0.852.Outcome summaries The Analysis_Outcomes sheet summarizes baseline and post-test sample sizes, means, standard deviations, mean gains, adjusted mean differences, 95% confidence intervals, and Hedges’ g for blending and segmentation.Qualitative coding Interview responses are coded by theme, valence, acceptability, ease of delivery, and barrier status.

Intervention teacher acceptability ratings are summarized on a 1–5 scale.ii. Temporal and geographical scope, temporal and spatial resolutionTemporal scopeThe workbook indicates the following study timeline:PhaseTemporal scope in workbookBaseline assessmentFebruary 3–7, 2025, inferred from envelope batch IDsIntervention implementation8 weeks, recorded as weeks 1–8Post-test assessmentApril 14–18, 2025, inferred from envelope batch IDsThe session log records intervention time by week number and planned session number, not by exact calendar date.

Fieldwork assessment records contain envelope batch IDs with embedded dates, which allow baseline and post-test windows to be identified.Temporal resolutionDifferent parts of the workbook have different temporal resolutions:Dataset componentTemporal resolutionLearner assessmentOne record per learner per assessment phase: baseline or post-testFieldwork logOne record per assessment eventSession logOne record per planned intervention sessionFidelity dataOne record per observed sessionTeacher interviewsOne coded row per interview question/response segmentAnalysis sheetsAggregated to feasibility indicator, outcome × arm, or fidelity indicatorGeographical scopeAll schools in the workbook are located in the Central Zone.

The sample includes both government-aided and low-fee private schools.Geographical / institutional unitCountSchools approached10Schools participating8Participating classrooms8Intervention classrooms4Comparison classrooms4The workbook does not include GPS coordinates, district names, parish names, or school addresses. Therefore, the spatial scope can be described only at the level of the Central Zone, school type, school ID, classroom ID, teacher ID, and learner ID.Spatial resolutionThe finest spatial unit in the workbook is the learner within a classroom within a school.

For implementation and fidelity data, the resolution is the intervention classroom/session. For school recruitment data, the resolution is the school.iii. Tabular data: number of entries, headers, and units of measurementOverall tabular structureThe workbook contains 15 sheets.

The main raw data sheets contain 1,037 primary records. Key headers and measurement unitsThe workbook uses the first row of each tabular sheet as the header row. The main headers and their meanings are summarized below.SheetImportant headersMeaning / unitsSchoolsschool id, region, school type, School response, allocated arm, eligible learners, consent and assent obtained, baseline completed, posttest completedSchool-level recruitment and participation counts.

Learner-related columns are counts.Teachers_Classroomsteacher id, classroom id, arm, grade, training status, supplement received, planned sessions if intervention, acceptability rating 1 to 5Classroom/teacher profile. Acceptability is a 1–5 rating. Planned sessions are counts.Learner_Rosterlearner id, arm, eligible, caregiver consent, child assent, baseline completed, posttest completed, missing post reason, gender, age years, assessor IDsLearner-level recruitment and participation.

Age is in years. Consent/completion fields are categorical Yes/No or not applicable.Fieldwork_Logfieldwork id, assessment id, learner id, phase, assessor id, admin language, individual quiet space, masking status, sealed envelopeAssessment administration and quality-control record. Mostly categorical fields.Assessment_Itemssubtask, item no, stimulus, target word, score rule, practice itemItem metadata for Luganda blending and segmentation.

Practice items are not scored.Assessment_Raw DataB1–B12, blending total, S1–S12, segmentation totalItem scores are binary 0/1. Totals are sums from 0 to 12.Session_Logweek, planned session no, game type, session status, dose counted, planned minutes, actual minutes, target group size, materials usedSession delivery data. Minutes are time units.

Target group size is learner count. Dose counted is 0/1.Fidelity_Raw Datamodelled I do, used familiar words, followed gradual release, gave individual turns, corrected errors, kept within 20 min, child engagement 1 to 5Fidelity indicators are binary 0/1. Engagement is a 1–5 rating.Teacher_Interview_Codestheme code, valence, short excerpt or summary, coded acceptability, coded ease of delivery, coded barrier, acceptability rating 1 to 5Qualitative coding.

Acceptability/ease are 1–5 ratings; barrier is 0/1.Analysis_Feasibilitynumerator_formula, denominator_formula, observed_value_formula, criterion, decision_formulaFeasibility statistics and decisions. Observed values are proportions or means.Analysis_Outcomesbaseline_n, baseline_mean, baseline_sd, posttest_n, posttest_mean, posttest_sd, mean_gain, adjusted_md, 95_ci, hedges_gOutcome summaries. Means/gains/adjusted MDs are score points on the 0–12 scale; Hedges’ g is standardized.Analysis_Fidelityobserved_count_formula, denominator_formula, mean_fidelity_formulaFidelity counts and proportions.Main outcome dataThe Assessment_Raw Data sheet has 314 assessment records:PhaseRecordsBaseline162Post-test152The assessment records are balanced by arm at each phase:ArmBaseline recordsPost-test recordsIntervention8176Comparison8176Outcome summaries in the workbook are:OutcomeArmBaseline meanPost-test meanMean gainBlendingIntervention4.38.54.2BlendingComparison4.16.32.2SegmentationIntervention3.87.73.9SegmentationComparison3.75.41.7The workbook also reports adjusted mean differences and standardized effects for intervention versus comparison:OutcomeReported adjusted mean difference95% CIHedges’ gBlending2.0[0.9, 3.1]0.59Segmentation2.1[1.0, 3.3]0.63iv.

Missing dataThe workbook contains missing cells, but most are clearly structural “not applicable” blanks.Recruitment and learner-level missingnessFrom 181 eligible learners:StatusCountCaregiver consent obtained171Child assent obtained168Baseline completed162Post-test completed152There are 19 learners without baseline assessment records. These are explained by:Baseline non-completion reasonCountNot in analytic sample13Ill on baseline day2Transferred before baseline2Absent during baseline visit2Among the 162 baseline learners, 10 did not complete post-test.

This attrition is balanced by arm: 5 intervention and 5 comparison. Reasons are:Post-baseline missing reasonCountSchool transfer4Prolonged absence4Illness on assessment day2Missing cells by sheetSheetMissingness patternSchools18 blanks. Agreed schools have no decline reason; declined schools have no randomized classroom or learner-count fields.Teachers_Classrooms4 blanks.

Acceptability rating is blank for comparison teachers because they did not receive the supplement.Learner_Roster200 blanks. These are mainly missing-post-reason blanks for learners who completed post-test, and assessor ID blanks for learners without baseline or post-test assessments.Fieldwork_LogNo blank cells found in the content range.Assessment_ItemsNo blank cells found.Assessment_Raw DataNo blank cells found.Session_Log78 blanks in “missed or shortened reason,” corresponding to delivered sessions where no reason was needed.Fidelity_Raw DataNo blank cells found.Teacher_Interview_Codes24 blanks.

Acceptability/ease/rating fields are blank for comparison-teacher rows where intervention acceptability was not applicable.Analysis_Outcomes6 blanks. Adjusted mean difference, 95% CI, and Hedges’ g are shown only on the intervention rows, not repeated on comparison rows.README / OverviewContain structural blank cells used for layout rather than missing data.No imputation method is documented or apparent in the workbook.v.

Data errors, error ranges, and likely causesFormula and range checksI found no Excel error values such as #REF!, #VALUE!, #DIV/0!, #NAME?, or #N/A in the workbook.The workbook contains 689 formula cells, mainly in:SheetFormula purposeAssessment_Raw DataSum B1–B12 into blending total and S1–S12 into segmentation totalAnalysis_FeasibilityCount and proportion formulas for feasibility indicatorsAnalysis_OutcomesCount formulas and gain formulasAnalysis_FidelityFidelity count, denominator, and mean formulasInternal consistency checks show:CheckResultBlending item scoresAll values are 0 or 1Segmentation item scoresAll values are 0 or 1Blending totalsValid range 0–12Segmentation totalsValid range 0–12Blending total formula vs item sums0 mismatchesSegmentation total formula vs item sums0 mismatchesFidelity indicatorsAll values are 0 or 1Child engagement ratingsValues fall within 3–5 on the 1–5 scaleActual session minutesValues fall within 0–20 minutes; 0 corresponds to missed sessionsDocumented data-quality issuesThe fieldwork and assessment sheets record 7 assessment events with possible assessor unmasking by child comment.

These same records indicate that the child referred to the games. This is a potential procedural limitation, not a missing-data problem.The main implementation challenges recorded in the session and fidelity notes include:Source of implementation difficultyEvidence in workbookTimetable pressure / late start8 missed or shortened reason entriesTimetable disruption5 entriesSchool event3 entriesTeacher illness2 entriesDrift toward chorus responsesObserver notes in fidelity dataNoise or uneven participationObserver notes in fidelity dataRushed or over-time sessionsObserver notes and “kept within 20 min” indicatorMinor inconsistency notedOne small inconsistency appears in the Analysis_Fidelity sheet: the raw fidelity count for “corrected errors” is 53/64 = 0.828125, which rounds to 0.828, but the reported target/value cell shows 0.831.

This appears to be a minor rounding or transcription inconsistency. It does not affect the overall fidelity calculation of 327/384 = 0.8515625, reported as 0.852.Error ranges and uncertaintyThe workbook does not provide measurement-error estimates, inter-rater reliability estimates, or standard errors for raw item scoring. The raw score scale itself constrains outcome values to 0–12 for each subtask.The workbook does provide statistical uncertainty for the adjusted outcome comparisons:OutcomeAdjusted mean difference95% confidence intervalBlending2.0 score points0.9 to 3.1Segmentation2.1 score points1.0 to 3.3Because the model specification for the adjusted mean differences is not included in the workbook, the reported adjusted effects can be described but not fully reproduced from the visible sheets alone.