# Action Research and Thesis Topics on Reading (CRLA) Twenty action research and thesis topics on reading for teachers and school heads, each with its question, design, test and the CRLA variables that answer it. This is the Markdown copy of [Action Research and Thesis Topics on Reading (CRLA)](https://nash.com.ph/blog/action-research-topics-reading-crla), an article on the NASH Reviewer Blog by Mark Anthony Llego, published Oct 28, 2026, updated Nov 4, 2026. Topics: Action research, Learning outcomes. NASH Reviewer is an independent practice platform not affiliated with DepEd, BHROD, or NEAP. > **Note** > > **What these data can and cannot do.** The cleaned CRLA data are school records: how many learners in each school reached each reading level in the Comprehensive Rapid Literacy Assessment at the end of SY 2025-26, joined to the school's SY 2023-24 teachers, classrooms and facilities. In DepEd's Action Research Toolkit, records like these are documents that help you *assess the situation*, the first stage of action research. They show where struggling readers are and what goes with them. They cannot show whether a reading strategy works: for that you assess your own learners before and after you try it. Each topic below says which kind of study it is. NASH Reviewer is not affiliated with DepEd or EDCOM II. Reading is the most common action research problem in Philippine elementary schools, and for good reason. In the end-of-year CRLA for SY 2025-26, 511,753 of 5,127,670 learners in Grades 1 to 3, 10.0%, were struggling readers, at the Low Emerging or High Emerging level, in 38,679 public schools. In Grade 1 Mother Tongue the share was 14.4%. Every one of those schools has a teacher who knows which learners they are. This post lists twenty research topics on reading in three groups: action research you can carry out in your own class or school, studies that assess the situation in a school or division, and topics for a master's thesis or doctoral dissertation. Each topic gives a research question, the CRLA variables that answer it (with their exact column names), a design and statistical test from DepEd's Action Research Toolkit, and a starting figure from the national data. Two worked examples then run the tests on the national data, and a section on pretests and posttests shows, with real CRLA results, why a reading intervention needs a comparison group. It is the companion of [Action Research and Thesis Topics on Classroom Shortage](https://nash.com.ph/blog/action-research-topics-classroom-shortage), written for Grades 1 to 3 teachers, reading coordinators, school heads, Learning Action Cell (LAC) groups and graduate students in education. Every file behind it is listed on the [EDCOM II CRLA data page](https://nash.com.ph/edcom-data). ## Key findings - DepEd defines action research as "a process of systematic and reflective inquiry that aims to improve educational practices or resolve problems in the classroom or in the school" (DepEd Order No. 16, s. 2017). When it tests a reading intervention, its design is experimental or qualitative, and you assess the learners yourself. - The CRLA data fit the first stage, assessing the situation, through causal-comparative, correlational and survey studies, and graduate theses that compare many schools. - Learners grow a great deal in one school year without any special intervention. In 6,061 schools in five regions, the share of Grade 1 learners struggling in Mother Tongue reading fell from 72.9% at the start of SY 2025-26 to 12.6% at the end. A pretest and a posttest alone cannot show that your intervention caused a gain; a comparison group can. - The unit of analysis changes the answer. Counting every learner, rural schools have more struggling readers than urban ones, 10.7% against 9.3%; counting every school once, the typical rural school has fewer, a median of 7.0% against 8.2%. - Crowded classrooms go with more struggling readers only weakly: Spearman's rho is 0.14 nationally and has a median of 0.04 within divisions, where it is below zero in 79 of 213 divisions. Most of the difference between schools lies elsewhere, which is where action research comes in. - A typical school assessed 71 learners in all and 23 in Grade 1, about one class. 6,173 schools assessed fewer than 30 learners, so one school's share can swing widely by chance. ## Action research, assessing the situation and theses: where the CRLA data fit DepEd's Research Management Guidelines (DepEd Order No. 16, s. 2017, amended by DepEd Order No. 26, s. 2021) define action research as "a process of *systematic* and *reflective inquiry* that aims to *improve* educational practices or *resolve* problems in the classroom or in the school". The Action Research Toolkit, which DepEd's Policy, Research and Development Division and Bureau of Learning Delivery prepared in 2019 with the Basic Education Sector Transformation (BEST) Program and the support of the Australian Government, describes five stages: assess the situation, ask a question, act to seek answers, acquire information, and analyze and reflect. The toolkit's own running example is a reading study: a teacher tests whether graphic organizers improve comprehension, with an experimental group that uses them and a control group that does not, a pretest for both at the start and a posttest at the end. It is also clear about designs: "There are only two possible research designs when interventions are applied—experimental and qualitative design." Causal-comparative, correlational and survey research are designs for assessing the situation, before you choose an intervention, and the toolkit lists data from DepEd's own database among the documents they can use. The CRLA data are records of this kind. | Kind of study | What it answers | Data you need | Topics below | |:--|:--|:--|:--| | Action research (experimental or qualitative) | Did my reading intervention help my learners? | Your own assessments before and after, with a comparison group where you can; the CRLA data as context | Action research topics | | Assessing the situation (causal-comparative, correlational, survey) | Where are the struggling readers, and what goes with them? | The CRLA data for your school, district or division; your own survey for the survey topic | Topics that assess the situation | | Thesis or dissertation | What explains the pattern across a region or the country? | The CRLA data with the school inputs, estimates and municipal data in the release | Thesis topics | The toolkit names three kinds of experiment. A pre-experiment has no comparison group, as in a single class with a pretest and a posttest; a quasi-experiment compares groups that were not formed at random, such as two sections; a true experiment assigns learners to groups at random. Pre-experiments and quasi-experiments are the most common in action research, because the work happens in the teacher's own classes. For reading, the section on pretests and posttests below explains why a quasi-experiment is worth the extra effort. ## Choosing a reading topic The toolkit asks you to rank possible topics on four criteria: *strategic importance* (how many other teachers or subjects would benefit), *urgency* (how soon a solution is needed), *magnitude* (how many learners would benefit) and *feasibility* (whether you can carry out the study with your time and resources). The CRLA data answer magnitude directly: your school's counts at each level tell you how many learners an intervention could reach. A good research question is clearly stated, worth answering for learners, ethical, feasible, and leads to an action. Reading topics fall under the teaching and learning theme of DepEd's Basic Education Research Agenda (DepEd Order No. 39, s. 2016), and topics on Indigenous Peoples learners and learners with difficulties also touch its cross-cutting area of inclusive education. The toolkit expects action research training to run through school [Learning Action Cell sessions](https://nash.com.ph/blog/deped-learning-action-cell). A Grades 1 to 3 LAC group is a natural team for a shared reading study: the teachers of parallel sections can serve as each other's comparison groups. The toolkit's reading list includes Nugent, Malik and Hollingsworth's *A Practical Guide to Action Research for Literacy Educators* (2012), a good next step for reading topics. ## Getting the CRLA data for your school and division - Download the [school estimates file](https://data.nash.com.ph/edcom-data/v1/school_estimates.csv.gz), which lists every school's share of struggling readers with its estimate and 95% interval; extract it with any zip tool and open it in Excel. - For reading levels by grade and language and the school inputs, download the [analysis table](https://data.nash.com.ph/edcom-data/v1/edcom_crla_2026.parquet), one row per school, and open it in Python, R or DuckDB. - Filter by `region` and `division`, then by `district` if your study covers a district, and find your school by its DepEd school ID in `school_id`. - Keep the columns your topic needs; the table below groups the main ones. - Write down the year of each variable: CRLA results are from the end of SY 2025-26, and school inputs from SY 2023-24, two years earlier. - Read a small school's share with care. [How to Read a School's CRLA Results Fairly](https://nash.com.ph/blog/how-to-read-school-crla-results) explains the estimates: the median 95% interval is 17 percentage points wide for schools with fewer than 30 learners. - Cite the data as shown at the end of this post, and name DepEd and the EDCOM II dashboard as the original sources. The table lists the main variables in the analysis table, what they measure, their year and how many of the 38,679 schools have a value. Each grade-and-language block has six columns: learners assessed (`_n`) and learners at Low Emerging (`_le`), High Emerging (`_he`), Developing (`_dev`), Transitioning (`_tra`) and At Grade Level (`_gl`). | Variables in edcom_crla_2026.parquet | What they measure | Year | Schools with a value | |:-----------------------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------------------------------------------------|:-----------|-----------------------:| | `school_id`, `school_name`, `region`, `division`, `district` | School identity and place | SY 2025-26 | 38,679 | | `g1_mt_n`, `g1_mt_le`, `g1_mt_he`, `g1_mt_dev`, `g1_mt_tra`, `g1_mt_gl` | Grade 1 Mother Tongue: learners by level | SY 2025-26 | 38,234 | | `g2_mt_*`, `g2_fil_*` | Grade 2 Mother Tongue and Filipino: learners by level | SY 2025-26 | 37,227 | | `g3_fil_*`, `g3_eng_*` | Grade 3 Filipino and English: learners by level | SY 2025-26 | 37,860 | | `total_n`, `struggling`, `total_gl` | Learners counted once (Grade 1 Mother Tongue, Grade 2 Filipino, Grade 3 English), struggling readers and readers at grade level | SY 2025-26 | 38,679 | | `language` | Language of the Mother Tongue assessment | SY 2025-26 | 38,679 | | `urban_rural`, `income_class`, `municipality_psgc`, `barangay_psgc` | Urban or rural, the LGU's income class, and PSA codes for joining other data | Various | 38,678 | | `latitude`, `longitude`, `coordinate_quality` | School location and a check on it | Various | 38,355 | | `head_type` | Highest school-leadership position | SY 2023-24 | 38,216 | | `in_enrol_es`, `in_enrol_g1`, `in_enrol_g2`, `in_enrol_g3`, `in_offers_multigrade`, `in_ip_share` | Enrollment, multigrade classes and Indigenous Peoples learners | SY 2023-24 | 38,216 | | `in_ptr_es`, `in_teachers_es`, `in_master_teachers` | Teachers and learners per teacher | SY 2023-24 | 38,213 | | `in_pcr_es`, `in_rooms_es`, `in_buildings_bad_share`, `in_makeshift_rooms` | Classrooms, learners per classroom and building condition | SY 2023-24 | 37,620 | | `in_elec_any_es`, `in_internet_es`, `in_water_es`, `in_no_functional_toilet`, `in_computers_per_100` | Electricity, internet, water, toilets and computers | SY 2023-24 | 38,216 | | `in_ellna_overall_mps`, `in_nat_overall_mps` | ELLNA Grade 3 and NAT Grade 6 school means | SY 2023-24 | 5,674 | **How many schools a division study has.** The 38,679 schools fall in 232 divisions. The median division has 57 schools, the middle half of divisions have 33 to 250, and they range from 9 to 1,111; 96 divisions have fewer than fifty. A school, in turn, is a small sample of learners: the median school assessed 71, and the median Grade 1 group 23. In an action research study of one or two sections, the toolkit's t-tests on your own learners' scores are the right tools; in a study of schools, compare schools with 30 or more learners, or use the estimates. ## Twenty research topics on reading | Topic | Kind of study | Design | |:--|:--|:--| | Moving Grade 1 Low Emerging readers | Action research | Quasi-experimental | | Grade 2 Filipino after Mother Tongue | Action research | Quasi-experimental | | Grade 3 English oral reading fluency | Action research | Quasi-experimental | | From Transitioning to grade level with graphic organizers | Action research | Quasi-experimental (the toolkit's example) | | Reading in multigrade classes | Action research | Pre-experimental or qualitative | | Reading in large classes | Action research | Quasi-experimental | | Reading at home with families | Action research | Pre-experimental and qualitative | | A school reading program led by the school head | Action research (school head) | Quasi-experimental, school level | | Urban and rural schools | Assessing the situation | Causal-comparative | | Schools with and without a principal | Assessing the situation | Causal-comparative | | Electricity, internet and reading | Assessing the situation | Causal-comparative | | Crowded classrooms and struggling readers | Assessing the situation | Correlational | | Grade 1 and Grade 3 in the same school | Assessing the situation | Correlational | | Teachers' reading practices in high-need schools | Assessing the situation | Survey | | School, division or region? | Thesis | Multilevel model | | School inputs and struggling readers | Thesis | Regression | | Where struggling readers cluster | Thesis | Spatial | | Stunting, day care and early reading | Thesis | Correlational, spatial | | CRLA, ELLNA and NAT in the same schools | Thesis | Correlational | | How reliable are school-reported CRLA results? | Thesis | Data quality study | ## Action research topics for your class or school These eight topics test an intervention. The CRLA data describe your school and division and give a baseline; the outcome data are yours to collect, with consent, before and after. The CRLA levels themselves make a convenient pretest and posttest, but the toolkit's tests apply to any reading measure you give consistently, such as an oral reading passage or a comprehension test. The toolkit teaches the paired t-test for one group measured twice and the independent t-test for two groups. ### Moving Grade 1 Low Emerging readers - **Research question:** Does a daily fifteen-minute letter-sound and decoding routine move more Grade 1 Low Emerging readers to a higher CRLA level than the usual lessons do? - **Variables from the data (context):** `g1_mt_n`, `g1_mt_le`, `g1_mt_he`, `g1_mt_dev`, `language` for your school and division. - **Data you collect:** CRLA levels (or a letter-sound test) at the start and end of the study for the learners in your section and in a comparison section; attendance. - **Design and analysis:** quasi-experimental, with a parallel section as the comparison group. Compare gains with an independent t-test, or compare the share of learners who moved up at least one level. - **Starting point in the data:** 9.4% of Grade 1 learners, 149,415 in all, were at Low Emerging in Mother Tongue reading at the end of SY 2025-26, and 14.4% were struggling readers, the highest share of any grade. ### Grade 2 Filipino after Mother Tongue - **Research question:** Does a bridging routine, reading the same story first in the Mother Tongue and then in Filipino, raise Grade 2 learners' Filipino reading levels? - **Variables from the data (context):** `g2_mt_*` and `g2_fil_*` for your school, and `language`. - **Data you collect:** Filipino and Mother Tongue reading levels before and after, for your section and a comparison section. - **Design and analysis:** quasi-experimental; independent t-test on gains, or a paired t-test within the group. - **Starting point in the data:** 7.9% of Grade 2 learners were struggling readers in Filipino, and nationally Grade 2 results in Filipino and in the Mother Tongue were nearly the same ([CRLA by Language](https://nash.com.ph/blog/crla-mother-tongue-filipino-english)). ### Grade 3 English oral reading fluency - **Research question:** Does repeated reading of short passages, three times a week, raise Grade 3 learners' English oral reading fluency? - **Variables from the data (context):** `g3_eng_*` and `g3_fil_*` for your school. - **Data you collect:** words read correctly per minute on graded passages before and after, and CRLA English levels. - **Design and analysis:** quasi-experimental; independent t-test on the gain in words per minute. - **Starting point in the data:** 8.1% of Grade 3 learners were struggling readers in English, against 5.6% in Filipino. Within a school the two go together closely (Spearman's rho 0.83), so a learner who struggles in English often needs help in Filipino too. ### From Transitioning to grade level with graphic organizers - **Research question:** Do graphic organizers improve the reading comprehension of Grade 3 learners at the Transitioning level? - **Variables from the data (context):** `g3_fil_tra`, `g3_fil_gl`, `g3_eng_tra`, `g3_eng_gl`, `total_tra`, `total_gl`. - **Data you collect:** comprehension test scores before and after, for an experimental and a control group of similar starting scores. - **Design and analysis:** the toolkit's own example: two groups, a pretest for both, the intervention for one, a posttest for both, and an independent t-test on the posttest or the gain. - **Starting point in the data:** 32.0% of learners were at the Transitioning level and 47.6% at grade level, so the step from Transitioning to grade level concerns the largest group of learners below it. ### Reading in multigrade classes - **Research question:** How does a cross-age reading buddy system, in which older learners read with younger ones, work in a multigrade class, and does it change the younger learners' reading levels? - **Variables from the data (context):** `in_offers_multigrade`, `in_multigrade_share_es`, `in_enrol_g1`, `in_enrol_g2`, `in_enrol_g3`. - **Data you collect:** reading levels before and after, observation notes and short interviews with learners. - **Design and analysis:** pre-experimental (one class, paired t-test) with a qualitative part; a multigrade class is often the only one of its kind in the school. - **Starting point in the data:** 5,828 schools offer multigrade classes. Their share of struggling readers is 11.8%, against 9.9% in other schools. ### Reading in large classes - **Research question:** Does a small-group rotation, in which the teacher reads with one group while the others work at stations, raise the reading levels of struggling readers in a class of more than forty? - **Variables from the data (context):** `in_pcr_es`, `in_ptr_es`, `in_rooms_es`; join to the [EDCOM II classroom data](https://nash.com.ph/edcom-classroom-data) by school ID for the 2023 classroom shortage and class shifts. - **Data you collect:** reading levels before and after for the struggling readers in your class and in a comparison class. - **Design and analysis:** quasi-experimental; independent t-test on gains. - **Starting point in the data:** crowding on its own goes with struggling readers only weakly (the second worked example below), which suggests that how a large class is organized may matter more than its size. ### Reading at home with families - **Research question:** Does a family reading log, with a short take-home reader each week, increase how often Grade 1 learners read at home, and do their reading levels rise? - **Variables from the data (context):** `language`, `in_ip_share`, `urban_rural`. - **Data you collect:** the reading logs, reading levels before and after, and interviews with parents. - **Design and analysis:** pre-experimental (paired t-test) with a qualitative part on what helped families take part. - **Starting point in the data:** in schools where Indigenous Peoples learners are half or more, 16.5% of learners were struggling readers, against 9.4% elsewhere; reading materials in the home language may be scarce there. ### A school reading program led by the school head - **Research question:** Does a school reading program, with CRLA-based grouping, weekly LAC sessions on reading and a daily reading block, reduce the school's share of struggling readers compared with similar schools in the district? - **Variables from the data (context):** `head_type`, `in_teachers_es`, `in_master_teachers`, `struggling`, `total_n`, and the same columns for the other schools in the district. - **Data you collect:** the school's CRLA results at the start and end of the year, the program's records, and teachers' reflections. - **Design and analysis:** quasi-experimental at the school level: compare the school's change with the change in similar schools, and add a qualitative account of how the program ran. - **Starting point in the data:** in the end-of-year CRLA, 65% of the variation between schools in the share of struggling readers sits between schools within a division ([CRLA Variation: School, Division or Region?](https://nash.com.ph/blog/crla-school-division-region-variation)). Schools in the same division differ a lot, which leaves room for what a school does. ## Pretest and posttest: why a reading study needs a comparison group The CRLA data include something few action researchers have: the same schools assessed at the start and at the end of one school year. In 6,061 public schools in five regions, the share of Grade 1 learners struggling in Mother Tongue reading fell from 72.9% to 12.6%, and 96.7% of schools raised their share of learners at grade level ([CRLA Start vs End of SY 2025-26](https://nash.com.ph/blog/crla-start-vs-end-of-year-sy-2025-2026)). ![Stacked bars of CRLA reading levels for five grade-and-language blocks at the start and end of SY 2025-26 in the same 6,061 schools. At grade level rises from 5% to 49% in Grade 1 Mother Tongue, from 17% to 51% in Grade 2 Filipino and from 23% to 53% in Grade 3 English, while the Low Emerging share in Grade 1 falls from 55% to 8%.](https://nash.com.ph/blog/_astro/crla-start-vs-end-of-year-levels.BgUg_g_3_1JeuAk.webp) *Figure 1. CRLA reading levels in each grade-and-language block at the start and at the end of SY 2025-26, in the same schools in five regions. Chart: © 2026 Mark Anthony Llego, NASH Reviewer. [Download the full-resolution chart (PNG)](https://nash.com.ph/blog/charts/deped-data/crla-start-vs-end-of-year-levels.png).* This is the first lesson for reading action research. Young readers grow a great deal in a year through ordinary teaching, so a single class that improves from pretest to posttest does not show that the intervention caused the gain. A paired t-test on one group will almost always find a gain. Compare your class's gain with a comparison section's, or at least with the typical gain in your school or district. The second lesson is regression to the mean. A class chosen because it scored lowest at the start is likely to rise more than others, partly because some of its low scores were bad luck on the day. In the same schools, the fifth of schools with the lowest share at grade level at the start rose by +42.1 points, against +29.8 for the highest fifth. ![Line chart of mean school share at grade level for five groups of schools ranked by their start-of-year share. The lowest group rises from 0% to 42% and the highest from 35% to 65%. Adjusting the start-of-year shares for school size moves the lowest group's start to 4% and the highest group's to 32%, so part of the gap between groups at the start is noise that disappears by the end.](https://nash.com.ph/blog/_astro/crla-regression-to-the-mean.DSp2TXzY_1730xs.webp) *Figure 2. Average school share of learners at grade level for five groups of schools ranked by their start-of-year share, at the start and at the end of SY 2025-26. Chart: © 2026 Mark Anthony Llego, NASH Reviewer. [Download the full-resolution chart (PNG)](https://nash.com.ph/blog/charts/deped-data/crla-regression-to-the-mean.png).* If you choose your struggling readers for an intervention, choose the comparison learners the same way, from the same starting levels, so that both groups rise for the same ordinary reasons and the difference between them is the intervention's. ## Topics that assess the situation in your school or division These six topics can be answered with the CRLA data alone, except the survey, which uses them to draw a sample. They compare groups that already differ, or relate variables, without an intervention: their results show associations, not causes. ### Urban and rural schools - **Research question:** Do rural schools in our division have a higher share of struggling readers than urban schools? - **Variables:** `urban_rural`, `struggling`, `total_n`. - **Design and analysis:** causal-comparative, comparing schools (Mann–Whitney test on school shares) and learners (pooled shares). - **Starting point in the data:** the [first worked example](#worked-example-urban-and-rural-schools-causal-comparative) shows that the answer depends on whether you count learners or schools. ### Schools with and without a principal - **Research question:** Do schools led by a School Principal have fewer struggling readers than schools led by a head teacher or with no head position, once school size and setting are taken into account? - **Variables:** `head_type`, `struggling`, `total_n`, `in_enrol_es`, `urban_rural`, `in_offers_multigrade`. - **Design and analysis:** causal-comparative, with matching or regression to compare schools of similar size and setting. - **Starting point in the data:** before any adjustment the gap looks large, 12.7% struggling readers in schools without a School Principal against 8.8%, but schools with a principal are much larger. Compared with similar schools, the difference is -1.2 percentage points (95% interval -1.6 to -0.8) ([CRLA Results in Schools With and Without a Principal](https://nash.com.ph/blog/schools-with-and-without-principal-crla)). ### Electricity, internet and reading - **Research question:** Do schools without electricity or internet have more struggling readers, and does the gap remain within the same division? - **Variables:** `in_elec_any_es`, `in_internet_es`, `struggling`, `total_n`, `division`. - **Design and analysis:** causal-comparative; compare pooled shares and school shares, then compare within divisions. - **Starting point in the data:** 13.7% of learners in schools without internet were struggling readers, against 9.4%, and 14.7% in the 917 schools without electricity, against 9.9%. Such schools are also smaller and more remote, so these raw gaps mix several things. ### Crowded classrooms and struggling readers - **Research question:** In our division, do schools with more learners per classroom have a higher share of struggling readers? - **Variables:** `in_pcr_es`, `in_ptr_es`, `struggling`, `total_n`. - **Design and analysis:** correlational: a scatterplot and Spearman's rho, read with the toolkit's guide. - **Starting point in the data:** the [second worked example](#worked-example-crowded-classrooms-and-struggling-readers-correlational) runs this in full: rho is 0.14 nationally. ### Grade 1 and Grade 3 in the same school - **Research question:** Do schools with many struggling readers in Grade 1 also have many in Grade 3? - **Variables:** `g1_mt_n`, `g1_mt_le`, `g1_mt_he`, `g3_eng_n`, `g3_eng_le`, `g3_eng_he` (or the `g3_fil_*` columns). - **Design and analysis:** correlational, across schools with 10 or more learners in each grade. - **Starting point in the data:** across 29,999 schools, Spearman's rho between the Grade 1 and Grade 3 shares of struggling readers is 0.47, a moderate link. These are different learners in the same year, so the result describes schools, not how a child's reading develops. ### Teachers' reading practices in high-need schools - **Research question:** How do Grades 1 to 3 teachers in schools with many struggling readers group learners, use the CRLA results and plan remediation, and what support do they ask for? - **Variables (to draw the sample):** in the school estimates file, `p_above_national`, `eb_rate`, `school_id`, `division`. - **Data you collect:** a short survey of mostly closed-ended items, with a few open questions. - **Design and analysis:** survey research. List the schools that are very likely above the national share, draw a random sample stratified by division, and survey a comparison sample from the other schools. - **Starting point in the data:** 8,280 schools have a probability above nineteen in twenty of an underlying share above the national one. ## Thesis and dissertation topics These six topics suit a master's thesis or a doctoral dissertation. They cover many schools, use several variables at once, and build on an earlier post in this series that gives a method and a starting result. ### School, division or region? - **Research question:** How much of the variation in struggling readers lies between regions, between divisions and between schools, and does it differ by grade and language? - **Variables:** `region`, `division`, `school_id`, and the counts for each grade-and-language block. - **Design and analysis:** a multilevel (hierarchical) binomial model. - **Starting point in the data:** 65% of the variation between schools sits between schools within a division, 22% between divisions and 13% between regions ([CRLA Variation: School, Division or Region?](https://nash.com.ph/blog/crla-school-division-region-variation)). ### School inputs and struggling readers - **Research question:** Which school inputs go with fewer struggling readers within the same division, once the others are held equal? - **Variables:** `in_ptr_es`, `in_pcr_es`, `in_buildings_bad_share`, `in_enrol_es`, `in_elec_any_es`, `in_internet_es`, `in_water_es`, `in_no_functional_toilet`, `in_offers_multigrade`, `in_ip_share`, `head_type`. - **Design and analysis:** a beta-binomial or binomial regression with division effects, reported as odds ratios with 95% intervals. - **Starting point in the data:** within a division, schools with more learners per teacher had more struggling readers, an odds ratio of 1.14 per standard deviation (95% interval 1.12 to 1.16), and schools with internet had fewer, an odds ratio of 0.89 ([School Resources and CRLA Reading Results](https://nash.com.ph/blog/school-resources-and-crla-reading)). ### Where struggling readers cluster - **Research question:** Do municipalities with many struggling readers cluster together, and where are the hot spots? - **Variables:** in the municipal table, `rate`, `eb_rate`, `lisa_q`, `gi_hot`, `adm3_pcode`; in the analysis table, `latitude`, `longitude`, `municipality_psgc`. - **Design and analysis:** spatial analysis: Moran's I, local indicators of spatial association and hot spot statistics. - **Starting point in the data:** Moran's I for the municipal shares is 0.65, and 64 municipalities form high–high clusters ([CRLA Reading Hotspots](https://nash.com.ph/blog/crla-reading-hotspots-map)). ### Stunting, day care and early reading - **Research question:** Do municipalities with more stunting among children aged 0–2, or more barangays without a child development center, have more struggling readers? - **Variables:** in the municipal table, `stunting`, `no_cdc_share`, `rate`, `n`. - **Design and analysis:** correlational at the municipal level, then a spatial regression. - **Starting point in the data:** across 1,639 municipalities and cities, the rank correlation with stunting is only 0.05 (95% interval -0.06 to 0.17) ([Stunting, Day Care Centers and Early Reading](https://nash.com.ph/blog/stunting-day-care-and-reading)). Municipal data cannot speak for individual children. ### CRLA, ELLNA and NAT in the same schools - **Research question:** Do schools with many struggling readers in the CRLA also have low ELLNA Grade 3 and NAT Grade 6 results? - **Variables:** `struggling`, `total_n`, `in_ellna_overall_mps`, `in_nat_overall_mps`, `in_ellna_n_test_takers`, `in_nat_n_test_takers`. - **Design and analysis:** correlational, weighting schools by the number of learners tested, with care for the different years and grades. - **Starting point in the data:** the ELLNA and NAT results come from fewer schools than the CRLA; [Do the Same DepEd Schools Lag on the ELLNA, NAT and CRLA?](https://nash.com.ph/blog/deped-schools-lag-ellna-nat-crla) compares the three. ### How reliable are school-reported CRLA results? - **Research question:** How often do school-reported CRLA results show patterns that chance rarely produces, and how much do they change the picture? - **Variables:** `total_n`, `struggling`, `total_gl`, the `g2_mt_*` and `g2_fil_*` columns, and the cleaning log. - **Design and analysis:** a data quality study: compare observed counts with those a beta-binomial model expects, and check for identical results across languages. - **Starting point in the data:** 8,543 schools report no struggling reader, close to the 7,490 expected when schools differ, and 14,765 schools report identical Grade 2 Mother Tongue and Filipino results level by level ([CRLA Schools With No Struggling Readers](https://nash.com.ph/blog/crla-schools-with-no-struggling-readers)). ## Worked example: urban and rural schools (causal-comparative) **Question.** Do rural public schools have a higher share of struggling readers than urban ones? **Hypotheses (nondirectional).** Null hypothesis: the share of struggling readers is the same in urban and rural schools. Research hypothesis: the shares differ. **Data.** `urban_rural`, `struggling` and `total_n` for every school with an urban or rural type. | Setting | Schools | Learners assessed | Struggling readers | Share struggling | Schools with 30 or more learners | Median school share | |:----------|----------:|--------------------:|---------------------:|:-------------------|-----------------------------------:|:----------------------| | Urban | 8,038 | 2,589,126 | 240,957 | 9.3% | 7,671 | 8.2% | | Rural | 30,504 | 2,504,027 | 268,731 | 10.7% | 24,716 | 7.0% | **Two ways to count.** Counting every learner, rural schools have more struggling readers: 10.7% against 9.3%. Counting every school once, among schools with 30 or more learners, the typical rural school has *fewer*: a median of 7.0% against 8.2%. The pooled share weighs each school by its learners, so large schools count for more; the median gives every school the same weight. Neither is wrong, but they answer different questions: one about learners, the other about schools. Decide which question you are asking before you analyze. **Test.** Learners in the same school are taught together and are not independent, so a chi-square test on millions of learners would greatly overstate how certain the difference is. Compare schools instead. A Mann–Whitney test on the school shares gives p < 0.001, with a rank-biserial correlation of -0.08: a rural school's share is higher than an urban school's in 46.0% of pairs. With so many schools even a tiny difference passes the test, so the effect size is what matters, and here it is close to nothing. **Interpretation.** Whether a school is urban or rural says little about its share of struggling readers. A rural school with many struggling readers should look for causes in its own situation, not in being rural. The chart compares pooled shares for five more characteristics. Each raw gap mixes the characteristic with everything that goes with it, such as size and remoteness; the topics above say how to separate them. ![Dot chart of the share of struggling readers among learners assessed in the CRLA at the end of SY 2025-26, in schools with and without six characteristics: rural (against urban) 10.7% against 9.3%; offers multigrade classes 11.8% against 9.9%; no school principal 12.7% against 8.8%; indigenous peoples learners are half or more 16.5% against 9.4%; no internet 13.7% against 9.4%; no electricity 14.7% against 9.9%.](https://nash.com.ph/blog/_astro/crla-struggling-readers-by-school-characteristic.CvOlQwRa_ZilaF8.webp) *Figure 3. Share of struggling readers among learners assessed in the CRLA at the end of SY 2025-26, in schools with and without each characteristic (SY 2023-24 school data). Chart: © 2026 Mark Anthony Llego, NASH Reviewer. [Download the full-resolution chart (PNG)](https://nash.com.ph/blog/charts/deped-data/crla-struggling-readers-by-school-characteristic.png).* | Schools with the characteristic | Schools | Share struggling | Share struggling in schools without it | Difference (points) | |:---------------------------------------------|----------:|:-------------------|:-----------------------------------------|----------------------:| | Rural (against urban) | 30,504 | 10.7% | 9.3% | 1.4 | | Offers multigrade classes | 5,828 | 11.8% | 9.9% | 2 | | No School Principal | 22,641 | 12.7% | 8.8% | 3.9 | | Indigenous Peoples learners are half or more | 5,946 | 16.5% | 9.4% | 7.1 | | No internet | 9,046 | 13.7% | 9.4% | 4.3 | | No electricity | 917 | 14.7% | 9.9% | 4.8 | ## Worked example: crowded classrooms and struggling readers (correlational) **Question.** Do schools with more learners per classroom have a higher share of struggling readers? **Data.** The 31,695 schools with 30 or more learners assessed and an elementary learner–classroom ratio, from `in_pcr_es` (SY 2023-24) and the school share of struggling readers (end of SY 2025-26). **Test.** Spearman's rho is 0.14 (95% interval 0.09 to 0.19, from a bootstrap that resamples whole divisions). On the toolkit's guide, where 0.1 to 0.3 is weak, 0.4 to 0.6 moderate and 0.7 to 0.9 strong, this is a weak, positive (direct) relationship. The table shows it in bands. | Elementary learners per classroom | Schools | Share struggling | Median school share | |:------------------------------------|----------:|:-------------------|:----------------------| | 30 or fewer | 18,724 | 9.3% | 6.4% | | Over 30 to 40 | 6,746 | 10.3% | 8.3% | | Over 40 to 50 | 2,612 | 10.9% | 8.8% | | Over 50 | 3,613 | 9.8% | 8.8% | **The same question, division by division.** A teacher or division researcher would run this test within one division. Computed separately in each of the 213 divisions with at least twenty such schools, rho has a median of 0.04 and runs from -0.55 to 0.55. It is below zero in 79 divisions, and reaches 0.4 or more in only 3. ![Histogram of Spearman's rho between elementary learners per classroom (SY 2023-24) and the share of struggling readers (end of SY 2025-26), computed within each of 213 divisions for schools with 30 or more learners assessed. The median is 0.04, values run from -0.55 to 0.55, 79 divisions are below zero, and the national value of 0.14 is marked.](https://nash.com.ph/blog/_astro/crla-crowding-correlation-by-division.DE0e3bYt_qMfNF.webp) *Figure 4. Spearman's rho between elementary learners per classroom (SY 2023-24) and the share of struggling readers (end of SY 2025-26), computed within each division for schools with 30 or more learners assessed, with the national value marked. Chart: © 2026 Mark Anthony Llego, NASH Reviewer. [Download the full-resolution chart (PNG)](https://nash.com.ph/blog/charts/deped-data/crla-crowding-correlation-by-division.png).* **Interpretation.** The national correlation mixes differences between divisions with differences within them; inside most divisions, crowding and struggling readers barely go together. A weak correlation is a finding too: it says that the number of learners per room does not by itself separate schools with many struggling readers from schools with few. Learners per teacher tell a similar story (rho 0.11). That is a reason to study what happens inside the classroom, which is what action research does. ## Common mistakes to avoid - **Crediting the intervention with ordinary growth.** Learners' reading improves over a year with regular teaching. Use a comparison group, or compare your gain with the usual gain. - **Choosing the lowest scorers and expecting a fair test.** They tend to rise anyway (regression to the mean). Choose comparison learners the same way. - **Counting learners when you mean schools, or the reverse.** Pooled shares and median school shares can point in opposite directions, as in the first worked example. - **Treating learners in the same school as independent.** Compare schools, or use multilevel models; a chi-square on learners overstates certainty. - **Ranking small schools by their raw share.** A school that assessed ten learners can show a very high or low share by chance. Use the estimates and their intervals. - **Reporting only p-values.** With thousands of schools nearly every difference passes a test. Report effect sizes and intervals. - **Reading school data as data about children.** The CRLA data here are counts per school. A link between schools does not show how any child is affected. - **Mixing years without saying so.** CRLA results are from the end of SY 2025-26; school inputs from SY 2023-24. - **Treating a screen as a test.** The CRLA is a teacher-administered screen; how and when it was given can differ between schools. In your own study, give the pretest and posttest the same way to every group. ## Ethics and approval The CRLA data are public school-level counts with no learner names, so using them needs no consent. Your own reading assessments are different. The toolkit sets five protocols for action research: reduce the risk of harm, secure informed consent (from parents, with the learners' assent), safeguard confidentiality and anonymity, avoid deceptive practices, and ensure justice and fidelity. For a reading study, justice means that the comparison group keeps its regular reading lessons and receives the intervention afterward if it works, so that no learner is held back for the sake of the study; fidelity means keeping your promises to learners and parents, such as their anonymity. Never post individual learners' reading levels, and report school results without ranking or shaming schools. Submit your proposal through your school and division under the Research Management Guidelines; action research proposals may be funded through the Basic Education Research Fund (BERF), whose rules are part of those guidelines (DepEd Order No. 16, s. 2017, as amended), so check your region's call each year. [BERF: How to Get DepEd Action Research Funded](https://nash.com.ph/blog/basic-education-research-fund-berf) explains the grant, and [DepEd Action Research Ethics: Consent, Assent and Privacy](https://nash.com.ph/blog/action-research-ethics-deped) covers consent forms and learners' data. ## What this means for school heads - **Start from your school's own counts.** Use the CRLA counts by grade and level to choose a reading problem with real magnitude, then check it against your own records. - **Plan studies with a comparison group.** Parallel sections in Grades 1 to 3 make natural comparison groups; schedule the pretest for every section and the intervention for one, with the others to follow. - **Use the LAC.** A Grades 1 to 3 LAC group can share one topic, one test and one analysis, and the results can feed the school's reading program and its [school improvement plan](https://nash.com.ph/blog/deped-school-improvement-plan). - **Compare with similar schools, not the national figure.** Your division's schools, and schools of similar size, are the fair comparison; the estimates file gives each school's interval. - **Read the data before acting on it.** [How to Read a School's CRLA Results Fairly](https://nash.com.ph/blog/how-to-read-school-crla-results) explains why small schools swing by chance. - **Write it up and share it.** The toolkit's final step is the report; a short report shared in the district lets other schools try what worked. ## How these numbers were made The data are the end-of-year CRLA for SY 2025-26 from the [EDCOM II early-childhood (ECCD) dashboard](https://edcom2.gov.ph/eccddashboard/) as of 10 May 2026, cleaned for this series and joined to DepEd's SY 2023-24 school files. The analysis was run in Python with pandas, NumPy and SciPy, and figures from earlier posts are linked where they appear. - **Learners.** Each learner is counted once: Grade 1 by the Mother Tongue result, Grade 2 by Filipino and Grade 3 by English. A struggling reader is at Low Emerging or High Emerging. - **Variables table.** For each group of columns, the schools with a value in its main column (for the CRLA blocks, at least one learner assessed). - **Worked example: urban and rural schools.** Pooled shares are struggling readers divided by learners assessed in each group. Median school shares and the Mann–Whitney test use schools with 30 or more learners assessed; the rank-biserial correlation is the probability that a rural school's share exceeds an urban school's, minus the reverse. - **Characteristics.** No School Principal compares schools whose highest position is a head teacher or no head position with schools with a School Principal. Indigenous Peoples learners are half or more when `in_ip_share` is at least one half. Electricity and internet are the elementary indicators. - **Worked example: crowded classrooms.** Schools with 30 or more learners assessed and an elementary learner–classroom ratio above zero. Spearman's rho with a 95% interval from 1,000 draws of whole divisions; within-division rho for divisions with at least twenty such schools. - **Grade 1 and Grade 3.** Schools with 10 or more learners assessed in both Grade 1 Mother Tongue and Grade 3 English. - **Start and end of the year.** The start-of-year results are DepEd's selected-school CRLA file, which covers five regions; the comparison uses the schools in both files. ## Limits The CRLA data describe schools, not learners, and one school year. The CRLA is a screen given by teachers, so differences in how it was given add to differences in reading. School inputs are from SY 2023-24 and may have changed by SY 2025-26. The start-of-year comparison covers five regions only. The toolkit's correlation guide leaves gaps between its bands; values below 0.4 are called weak here. The national results in the worked examples are illustrations, and a division's results may differ, as the second example shows. All results are associations. The Action Research Toolkit quoted here is the 2019 edition. DepEd may update its research guidelines, agenda and funding rules, and the CRLA's levels and grades; check the current issuances with your division's research staff. ## Frequently asked questions ### What are good action research topics on reading? Topics that test something you can change in your own class: a daily decoding routine for Grade 1 Low Emerging readers, a bridging routine from the Mother Tongue to Filipino, repeated reading for Grade 3 English fluency, graphic organizers for comprehension, reading buddies in a multigrade class, small-group rotation in a large class, or a family reading log. The CRLA data describe the setting; you assess your learners before and after. ### Can I use the CRLA results as my pretest and posttest? Yes, if you give the CRLA the same way to every group, but add a comparison group. Learners' CRLA levels rise a great deal over a school year with ordinary teaching, so a gain in one class alone does not show that your intervention caused it. Compare your class's gain with a parallel section's. ### Which variables measure struggling readers in the CRLA data? `struggling` counts learners at Low Emerging or High Emerging, out of `total_n` learners assessed, counting each learner once. For each grade and language, the `_le`, `_he`, `_dev`, `_tra` and `_gl` columns give the learners at each level, out of the block's `_n`. The school estimates file adds each school's estimate, 95% interval and probability above the national share. ### What statistical test should I use for a reading action research? For one class with a pretest and a posttest, a paired t-test; for an experimental and a comparison group, an independent t-test on the posttest or the gain, as the toolkit teaches. For comparing groups of schools, a Mann–Whitney test on school shares; for two measures across schools, Spearman's rho with a scatterplot; for several predictors, a regression with division effects. ## References - Creswell, J. W. (2012). *Educational research: Planning, conducting, and evaluating quantitative and qualitative research* (fourth edition). Pearson. - Department of Education. (2016). *Adoption of the Basic Education Research Agenda* (DepEd Order No. 39, s. 2016). - Department of Education. (2017). *Research management guidelines* (DepEd Order No. 16, s. 2017). - Department of Education. (2021). *Amendment to DepEd Order No. 16, s. 2017 (Research management guidelines)* (DepEd Order No. 26, s. 2021). - Department of Education & Basic Education Sector Transformation (BEST) Program. (2019). *Action research toolkit: Teacher's guide*. Department of Education, with support from the Australian Government. - Nugent, G., Malik, S., & Hollingsworth, S. (2012). *A practical guide to action research for literacy educators*. International Reading Association. ## Download the data The cleaned EDCOM II dashboard data behind this series are free to download for personal study; to reuse them, email [support@nash.com.ph](mailto:support@nash.com.ph) for written permission first. [How the EDCOM Dashboard Data Was Cleaned](https://nash.com.ph/blog/edcom-eccd-dashboard-data-cleaned) explains every table and cleaning rule. The files are at data.nash.com.ph: start with the [README](https://data.nash.com.ph/edcom-data/v1/README.md), then the [school table with CRLA results and school inputs](https://data.nash.com.ph/edcom-data/v1/edcom_crla_2026.parquet) (Parquet), the [school estimates](https://data.nash.com.ph/edcom-data/v1/school_estimates.csv.gz) (CSV), the [division estimates](https://data.nash.com.ph/edcom-data/v1/division_estimates.csv) (CSV) and the [municipal table](https://data.nash.com.ph/edcom-data/v1/municipalities.csv) (CSV). The DepEd school files used for school inputs are described in [Download Cleaned DepEd School Data](https://nash.com.ph/blog/download-cleaned-deped-school-data). Every file, with its rows, columns and checksum, is listed on the [EDCOM II CRLA data page](https://nash.com.ph/edcom-data). To cite the data: Llego, M. A. (2026). *EDCOM II early-childhood dashboard data, cleaned, version 1*. NASH Reviewer. https://nash.com.ph/edcom-data ## More in this series - [CRLA End-of-Year Results SY 2025-26: All 18 Regions](https://nash.com.ph/blog/crla-end-of-year-results-sy-2025-2026) - [CRLA Start vs End of SY 2025-26 in the Same Schools](https://nash.com.ph/blog/crla-start-vs-end-of-year-sy-2025-2026) - [CRLA by Language: Mother Tongue, Filipino and English](https://nash.com.ph/blog/crla-mother-tongue-filipino-english) - [How to Read a School's CRLA Results Fairly](https://nash.com.ph/blog/how-to-read-school-crla-results) - [CRLA Variation: School, Division or Region?](https://nash.com.ph/blog/crla-school-division-region-variation) - [Where Struggling Readers Are Concentrated](https://nash.com.ph/blog/where-struggling-readers-are-concentrated) - [School Resources and CRLA Reading Results](https://nash.com.ph/blog/school-resources-and-crla-reading) - [CRLA Results in Schools With and Without a Principal](https://nash.com.ph/blog/schools-with-and-without-principal-crla) - [Action Research and Thesis Topics on Reading (CRLA)](https://nash.com.ph/blog/action-research-topics-reading-crla) - [CRLA Reading Hotspots: A Map of Struggling Readers](https://nash.com.ph/blog/crla-reading-hotspots-map) - [Stunting, Day Care Centers and Early Reading](https://nash.com.ph/blog/stunting-day-care-and-reading) - [Literacy and Nutrition Programs Compared With Need](https://nash.com.ph/blog/literacy-nutrition-programs-coverage) - [CRLA Schools With No Struggling Readers: A Data Check](https://nash.com.ph/blog/crla-schools-with-no-struggling-readers) - [How the EDCOM Dashboard Data Was Cleaned](https://nash.com.ph/blog/edcom-eccd-dashboard-data-cleaned) - Regional reports: [NCR](https://nash.com.ph/blog/crla-results-sy-2025-2026-ncr), [CAR](https://nash.com.ph/blog/crla-results-sy-2025-2026-car), [Region I](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-1-ilocos), [Region II](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-2-cagayan-valley), [Region III](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-3-central-luzon), [Region IV-A](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-4a-calabarzon), [MIMAROPA](https://nash.com.ph/blog/crla-results-sy-2025-2026-mimaropa), [Region V](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-5-bicol), [Region VI](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-6-western-visayas), [NIR](https://nash.com.ph/blog/crla-results-sy-2025-2026-negros-island-region), [Region VII](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-7-central-visayas), [Region VIII](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-8-eastern-visayas), [Region IX](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-9-zamboanga-peninsula), [Region X](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-10-northern-mindanao), [Region XI](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-11-davao), [Region XII](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-12-soccsksargen), [Caraga](https://nash.com.ph/blog/crla-results-sy-2025-2026-region-13-caraga), [BARMM](https://nash.com.ph/blog/crla-results-sy-2025-2026-barmm) - Earlier reports: [CRLA Results SY 2025-2026 (DepEd's selected-school file)](https://nash.com.ph/blog/crla-results-sy-2025-2026), [Do the Same DepEd Schools Lag on the ELLNA, NAT and CRLA?](https://nash.com.ph/blog/deped-schools-lag-ellna-nat-crla), [DepEd School Inputs vs Test Results](https://nash.com.ph/blog/deped-school-inputs-vs-test-results), [EDCOM II Final Report: Guide for Aspiring School Heads](https://nash.com.ph/blog/edcom-ii-final-report), [DepEd Data for Aspiring School Heads: Where to Start](https://nash.com.ph/blog/deped-data-aspiring-school-heads) *Analysis and charts by Mark Anthony Llego for NASH Reviewer. Data: EDCOM II early-childhood dashboard and Department of Education public datasets. NASH Reviewer is not affiliated with EDCOM II or DepEd. Charts may be reused only with written permission from support@nash.com.ph.* --- © 2026 Mark Anthony Llego, NASH Reviewer. All rights reserved. You may quote up to 100 words from this article with the author's name, its title and a link. In research, cite it as: Llego, M. A. (2026, October 28). *Action Research and Thesis Topics on Reading (CRLA)*. NASH Reviewer. https://nash.com.ph/blog/action-research-topics-reading-crla. Any other reuse of the article, its charts or its data needs written permission: email support@nash.com.ph. Terms of Service: https://nash.com.ph/terms#website-content-and-data