# Cluster Sampling in Field Biology ## Quick Answer - Cluster sampling reduces field costs by sampling groups of organisms instead of individuals across a large landscape, but it requires statistical correction for within-group similarity. - Select clusters using probability methods and measure intra-cluster correlation to calculate the effective sample size before data collection begins. - The main limitation is that clusters that are similar to each other but internally homogeneous can produce biased estimates if the number of clusters is too small. ## What Is Cluster Sampling in Field Biology Cluster sampling is a probability-based sampling design in which the researcher divides a study area into discrete groups, called clusters, and then randomly selects a subset of those clusters for measurement. Within each selected cluster, the researcher measures either all individuals or a random subsample of individuals. This design contrasts with simple random sampling, where every individual in the study area has an equal chance of being selected, and with stratified sampling, where the study area is divided into homogeneous subgroups and random samples are drawn from each subgroup. The defining feature of cluster sampling is that the sampling unit is not the individual organism but the group. For example, a researcher studying plant diversity in a forest might divide the forest into 100 one-hectare plots, randomly select 20 plots, and then count every tree species within each selected plot. The plot is the cluster, and the trees are the individual units. This approach reduces the number of locations the researcher must visit, which lowers travel time, equipment setup, and personnel costs. Cluster sampling is most useful when a complete list of individual organisms is unavailable or impractical to construct. In many field settings, researchers can map the boundaries of a study area and divide it into grid cells or habitat patches more easily than they can enumerate every animal or plant. The design also supports studies where the natural grouping of organisms matters, such as social groups of animals, nests in a colony, or plants within a single soil patch. The primary tradeoff is statistical efficiency. Individuals within a cluster tend to be more similar to each other than individuals in different clusters. This similarity, called intra-cluster correlation, reduces the amount of independent information each measured individual provides. A sample of 500 individuals drawn from 10 clusters may provide less statistical power than a sample of 500 individuals drawn from 100 clusters, even if the total number of measured individuals is identical. ## At a Glance | Design Feature | Cluster Sampling | Simple Random Sampling | Stratified Sampling | | --- | --- | --- | --- | | Sampling unit | Group or plot | Individual organism | Individual within stratum | | Cost and logistics | Lower travel and setup cost | Higher cost for large areas | Moderate cost | | Statistical efficiency | Lower when intra-cluster correlation is high | Highest per individual | High per individual | | Best use case | Large areas with natural grouping | Small areas with complete list | Heterogeneous habitat with known strata | | Key risk | Correlation inflates precision | Missing rare subpopulations | Requires accurate stratum boundaries | ## Core Principles of Cluster Sampling ### The Cluster as the Primary Sampling Unit The cluster is the primary sampling unit in this design. The researcher first defines the cluster boundaries, then selects clusters, and finally measures individuals within the selected clusters. The number of clusters selected is the sample size that determines statistical power, not the number of individual organisms measured. This distinction is critical for planning because it changes how the researcher calculates the required sample size. For example, a researcher studying amphibian populations in a wetland complex might define each wetland as a cluster. If the researcher selects 15 wetlands and measures 30 frogs per wetland, the total sample is 450 frogs, but the effective sample size is closer to 15 because the wetland is the independent unit. The 30 frogs within a wetland are not independent observations because they share the same habitat conditions, water chemistry, and predator community. The researcher must decide whether the cluster is the natural unit of interest or a convenience unit. If the research question concerns individual-level variation, such as body condition or disease prevalence, the cluster is a convenience unit that reduces cost. If the research question concerns cluster-level variation, such as habitat quality or population density, then the cluster is the natural unit and the analysis should focus on cluster-level summaries. ### Intra-Cluster Correlation Intra-cluster correlation is the statistical measure of how similar individuals are within a cluster compared to individuals in different clusters. The value ranges from zero to one. A value of zero means individuals within a cluster are no more similar than individuals across clusters, and the design provides the same statistical efficiency as simple random sampling. A value close to one means individuals within a cluster are nearly identical, and the design provides little information beyond the number of clusters. The intra-cluster correlation coefficient is estimated from pilot data or from published values for similar species and habitats. It is used to calculate the design effect, which is the ratio of the variance under cluster sampling to the variance under simple random sampling. The design effect is then used to adjust the sample size calculation. A researcher who ignores intra-cluster correlation will underestimate the number of clusters needed and will report confidence intervals that are too narrow. ### Cluster Size and Number of Clusters The number of clusters and the size of each cluster are the two decisions that determine the cost and precision of the study. Increasing the number of clusters increases precision but also increases travel and setup cost. Increasing the size of each cluster increases the number of measured individuals but does not increase precision as much as adding a new cluster, because the additional individuals are correlated with the existing individuals in the cluster. The researcher must balance the number of clusters against the cluster size. A common rule is to maximize the number of clusters and minimize the cluster size, because the number of clusters is the primary determinant of statistical power. However, the cluster size must be large enough to provide a stable estimate of the cluster mean and to capture the within-cluster variation that is part of the research question. ## Comparing Cluster Sampling with Other Field Methods ### Cluster Sampling versus Simple Random Sampling Simple random sampling requires a complete list of all individuals in the study area. This list is often impossible to construct for mobile or cryptic organisms. Cluster sampling requires only a list of clusters, which is easier to construct because clusters are defined by geographic boundaries or habitat patches. Simple random sampling provides the highest statistical efficiency per individual because every individual is an independent observation. Cluster sampling provides lower efficiency because individuals within a cluster are correlated. The efficiency loss is acceptable when the cost of reaching an individual is high and the cost of measuring an additional individual within a cluster is low. The researcher should choose simple random sampling when the study area is small, the population is fully enumerated, and the cost of reaching each individual is similar. The researcher should choose cluster sampling when the study area is large, the population is not enumerated, and the cost of reaching a new location is high. ### Cluster Sampling versus Stratified Sampling Stratified sampling divides the population into homogeneous subgroups and samples from each subgroup. The purpose is to ensure that each subgroup is represented in the sample. Cluster sampling divides the population into heterogeneous groups and samples a subset of the groups. The purpose is to reduce cost by visiting fewer locations. The two designs are sometimes confused because both involve dividing the study area. The key difference is the goal. Stratification is used to reduce variance by ensuring coverage of known subgroups. Clustering is used to reduce cost by accepting some variance increase. A researcher can combine the two designs by stratifying the study area into habitat types and then cluster sampling within each stratum. ### Cluster Sampling versus Systematic Sampling Systematic sampling selects individuals at a fixed interval across the study area. For example, the researcher might measure every tenth plant along a transect. Systematic sampling is easy to implement and ensures even coverage of the study area. It is a form of cluster sampling when the interval creates groups, but it is usually treated as a separate design. The main risk of systematic sampling is that the interval may coincide with a natural pattern in the study area. If the interval matches the spacing of a habitat feature, the sample will be biased. Cluster sampling avoids this risk by using random selection of clusters, but it may produce uneven coverage of the study area. ## Practical Workflow for Cluster Sampling ### Step 1 Define the Study Area and the Cluster Boundaries The researcher must first define the geographic extent of the study area and the boundaries of the clusters. The cluster boundaries must be unambiguous and mappable. The researcher should use a map or a geographic information system to draw the boundaries and to calculate the area of each cluster. The cluster boundaries should be based on natural or management-relevant features. For example, a researcher studying grassland birds might define clusters as individual fields or pastures. A researcher studying stream invertebrates might define clusters as 100-meter stream reaches. The boundaries must be stable for the duration of the study so that the researcher can return to the same cluster for repeated measurements. ### Step 2: Construct the Cluster List The researcher must create a complete list of all clusters in the study area. This list is the sampling frame. The list must include every cluster that could be selected, and each cluster must have a known probability of selection. If the list is incomplete, the sample will be biased because some parts of the study area will have no chance of being selected. The cluster list can be constructed from maps, satellite imagery, land records, or field reconnaissance. The researcher should verify the list by walking or driving the study area to confirm that every cluster is present and that no cluster has been missed. The list should be recorded in a spreadsheet or a geographic information system with a unique identifier for each cluster. ### Step 3: Select the Clusters The researcher selects the clusters using a random or probability-based method. Simple random selection gives each cluster an equal chance of being selected. Systematic selection, such as selecting every fifth cluster from a random starting point, is also acceptable but must be checked for alignment with natural patterns. The number of clusters to select is determined by the sample size calculation, which accounts for the intra-cluster correlation and the desired precision. The researcher should select more clusters than the minimum calculation suggests to allow for loss of clusters due to access problems, landowner refusal, or weather damage. ### Step 4: Measure the Individuals within the Selected Clusters The researcher must decide whether to measure all individuals within the selected cluster or a subsample. Measuring all individuals is called a census of the cluster. Measuring a subsample is called two-stage cluster sampling. The subsample is selected using a second random process within the cluster. The measurement protocol must be identical across all clusters. The researcher should write a field protocol that specifies the measurement method, the order of measurement, the equipment, and the data recording format. The protocol should be tested in a pilot cluster before the full study begins. ### Step 5: Record the Data and the Cluster Identifiers Each individual measurement must be linked to its cluster identifier. The researcher must record the cluster identifier, the date, the observer, and the measurement values. The data should be entered into a spreadsheet or database with a column for the cluster identifier so that the analysis can account for the clustering. The researcher should also record the cluster-level variables, such as the cluster area, the habitat type, and the number of individuals measured. These variables are used in the analysis to explain the variation between clusters. ## Records and Measurements The researcher must keep a field notebook and a digital record of the sampling process. The field notebook should contain the cluster map, the cluster list, the random selection record, and the field protocol. The digital record should contain the raw data, the cluster identifiers, and the cluster-level variables. The researcher should record the following measurements for each cluster: - The cluster identifier and the geographic coordinates of the cluster center - The date and time of the sampling visit - The number of individuals measured in the cluster - The total number of individuals estimated in the cluster, if a census was not possible - The habitat type and any environmental variables measured at the cluster The researcher should also record the time spent in each cluster and the travel time between clusters. This information is used to calculate the cost of the sampling design and to plan future studies. ## Calculating the Number of Clusters The number of clusters is the primary determinant of the statistical power of the study. The researcher must calculate the number of clusters before the field season begins. The calculation requires the following inputs: - The desired level of precision, expressed as the width of the confidence interval - The expected variance of the response variable - The intra-cluster correlation coefficient - The cluster size The design effect is calculated from the intra-cluster correlation and the cluster size. The design effect is then multiplied by the sample size that would be required under simple random sampling. The result is the number of individuals needed under cluster sampling. The number of clusters is then calculated by dividing the adjusted sample size by the cluster size. The researcher should use a statistical software package or a sample size calculator to perform this calculation. The calculation should be repeated for a range of plausible intra-cluster correlation values to test the sensitivity of the design. If the number of clusters is too large for the budget, the researcher must either accept a lower precision or change the cluster size. ## Common Failure Patterns in Cluster Sampling ### Failure to Account for Intra-Cluster Correlation The most common failure is to analyze the data as if the individuals were independent. This produces confidence intervals that are too narrow and p-values that are too small. The researcher may conclude that a difference is statistically significant when the difference is not supported by the data. The researcher must use a statistical method that accounts for the cluster, such as a mixed-effects model with the cluster as a random effect, or a generalized estimating equation. The analysis must include the cluster identifier as a variable. ### Selecting Clusters That Are Too Similar If the researcher selects clusters that are all in the same habitat type or the same geographic region, the sample will not represent the study area. The researcher must ensure that the cluster list includes the full range of habitat types and geographic locations. The random selection process will then produce a representative sample. ### Using Too Few Clusters A study with only a few clusters has low statistical power and cannot detect small effects. The researcher must calculate the number of clusters and must not reduce the number to save cost. A study with fewer than the calculated number of clusters should be reported as a pilot study or a descriptive study, not as a hypothesis test. ### Cluster Boundaries That Change If the cluster boundaries change during the study, the data cannot be compared across time. The researcher must mark the cluster boundaries in the field and must confirm that the boundaries are stable. If the boundaries change, the researcher must document the change and must treat the changed cluster as a new cluster. ## Quality Controls and Field Checks The researcher should implement quality controls to ensure that the data are accurate and that the sampling design is followed. The quality controls include the following: - A pilot study to test the field protocol and to estimate the intra-cluster correlation - A second observer to repeat a subset of the measurements to estimate the measurement error - A check of the cluster list against the map to confirm that no cluster was missed - A check of the data entry to confirm that the cluster identifiers are correct - A review of the data to identify any clusters with unusual values that may indicate a field error The researcher should also consider the welfare of the organisms being studied. The sampling protocol should minimize the disturbance to the organisms and their habitat. The researcher should follow the institutional animal care and use guidelines and should obtain the required permits before the study begins. ## Professional Escalation Criteria The researcher should seek professional advice when the study design is complex or when the data analysis requires specialized statistical methods. The researcher should consult a statistician when: - The intra-cluster correlation is unknown and cannot be estimated from prior data - The cluster size varies widely across the clusters - The study requires a complex design with multiple levels of clustering - The data analysis requires a mixed-effects model or a generalized estimating equation The researcher should also consult the institutional review board or the animal care and use committee when the study involves animals or when the sampling may affect the population. ## Decision Framework for Cluster Selection and Budget Allocation ### The Cluster Selection Decision Matrix Field biologists often struggle with the practical question of which clusters to select when the study area contains clusters of different sizes, accessibility levels, and habitat qualities. A structured decision matrix helps researchers move from intuition to a defensible selection process. The matrix should be completed before any field visits begin and should be reviewed by the research team to ensure that the selection criteria are transparent and reproducible. The decision matrix begins with the cluster list that was constructed during the workflow. For each cluster, the researcher records the following attributes: cluster area, estimated population size, habitat type, accessibility score, landowner permission status, and any known risks such as flooding or seasonal inaccessibility. The accessibility score is a simple ordinal rating from one to five, where one means the cluster is reachable by vehicle with no walking required and five means the cluster requires a multi-day hike or specialized equipment. This score is recorded before selection so that the researcher can assess whether the final cluster set is biased toward accessible locations. The selection process then proceeds through three filters. The first filter removes clusters that are not viable for the study. A cluster is not viable if the landowner refuses access, if the cluster has been destroyed by recent disturbance, or if the cluster cannot be safely reached during the planned sampling window. The researcher must document the reason for each removal so that the final sample can be described accurately in the methods section. The second filter applies the probability selection method. The researcher assigns each viable cluster a selection probability based on the study design. In equal probability selection, every viable cluster has the same chance of being chosen. In probability proportional to size selection, larger clusters have a higher chance of being chosen. The second approach is often more efficient because it reduces the variance of the population estimate, but it requires that the cluster size be known before selection. The third filter is the balance check. After the random selection is complete, the researcher compares the selected clusters to the full cluster list for the key attributes. If the selected clusters are all in one habitat type or all in one geographic region, the researcher should consider whether the random draw produced an unrepresentative sample. The researcher may draw a new random selection or may add clusters to improve balance, but any deviation from the original random selection must be documented and reported. ### The Cost Efficiency Decision Rule The cost efficiency decision rule helps the researcher determine whether cluster sampling is the right design for the study and how to allocate the budget between the number of clusters and the cluster size. The rule is based on the relative cost of reaching a new cluster versus the cost of measuring an additional individual within a cluster. The researcher first estimates the cost of reaching a new cluster, which includes travel time, fuel, vehicle maintenance, and the time spent navigating to the cluster. The researcher then estimates the cost of measuring an additional individual within a cluster, which includes the time per measurement, the equipment wear, and the personnel time. The ratio of these two costs determines the optimal cluster size. If the cost of reaching a new cluster is high and the cost of measuring an additional individual is low, the researcher should select a small number of clusters and measure many individuals within each cluster. This situation occurs in remote field sites where travel between clusters requires days. If the cost of reaching a new cluster is low and the cost of measuring an additional individual is high, the researcher should select a large number of clusters and measure few individuals within each cluster. This situation occurs in accessible study areas where the measurement itself is time-consuming, such as behavioral observations or tissue sampling. The researcher should calculate the cost ratio before the field season and should use the ratio to set the cluster size. The calculation is recorded in the field notebook so that the decision can be reviewed and defended. The cost ratio should be recalculated after the pilot study because the pilot study provides the actual time measurements for both travel and individual measurement. ### The Cluster Size Adjustment Protocol The cluster size adjustment protocol is used when the researcher discovers during the field season that the planned cluster size is not achievable. This situation occurs when the cluster contains fewer individuals than expected, when the individuals are more difficult to find than expected, or when the field season is shortened by weather or access problems. The protocol begins with a decision rule. If the researcher planned to measure 30 individuals per cluster but the cluster contains only 20 individuals, the researcher must decide whether to measure all 20 individuals or to move to the next cluster. The decision depends on the research question. If the research question concerns the cluster mean, the researcher should measure all available individuals because the cluster mean is more stable with a larger sample. If the research question concerns the population total, the researcher should measure the available individuals and record the cluster size so that the analysis can weight the cluster appropriately. The protocol also includes a rule for the minimum cluster size. The researcher sets a minimum number of individuals per cluster before the field season begins. If the cluster contains fewer than the minimum, the cluster is recorded as a partial cluster and the data are flagged in the dataset. The researcher must not replace the partial cluster with a different cluster because the replacement would bias the sample toward clusters that are easier to sample. The researcher should also record the reason for the partial cluster. The reason is recorded in the field notebook and in the digital dataset. The reason is used in the analysis to determine whether the partial clusters are systematically different from the full clusters. If the partial clusters are all in one habitat type, the analysis must account for the missing data. ### The Cluster Replacement Protocol The cluster replacement protocol is used when a selected cluster cannot be sampled after the field season begins. The cluster may be inaccessible because of flooding, the landowner may revoke permission, or the cluster may be destroyed by a natural event. The researcher must have a replacement protocol in place before the field season begins. The protocol specifies that the researcher must not replace the lost cluster with a cluster that is convenient or that is known to be similar to the lost cluster. The replacement must be selected from the original cluster list using the same probability method that was used for the initial selection. The researcher must record the replacement and the reason for the replacement in the field notebook. The protocol also specifies the timing of the replacement. The researcher should replace the lost cluster as soon as the loss is known so that the replacement can be sampled within the same season. If the replacement cannot be sampled within the same season, the researcher must decide whether to continue with the reduced number of clusters or to extend the field season. The decision is recorded in the field notebook and is reported in the methods section. The researcher must also consider the effect of the replacement on the intra-cluster correlation estimate. If the replacement cluster is different from the lost cluster, the intra-cluster correlation may change. The researcher should recalculate the design effect after the replacement and should report the final number of clusters and the final intra-cluster correlation in the results. ### The Cluster Sampling Record System The record system for cluster sampling is a structured set of documents that the researcher maintains throughout the study. The system includes the cluster list, the selection record, the field visit log, the measurement record, and the analysis record. Each record is linked to the cluster identifier so that the researcher can trace the data from the field to the final analysis. The cluster list is the master record of all clusters in the study area. The list includes the cluster ID, the geographic coordinates, the cluster area, the habitat type, and the selection probability. The list is created during the workflow step and is updated when clusters are removed or replaced. The selection record is the documentation of the random selection process. The record includes the random number seed, the selection method, the list of selected clusters, and the list of replacement clusters. The selection record is used to verify that the selection was random and to reproduce the selection if the study is audited. The cluster visit log is the record of each field visit to a cluster. The log includes the date, the time, the observer, the weather conditions, the number of individuals measured, and any problems encountered. The log is completed in the field and is entered into the digital record system at the end of each day. The measurement record is the raw data from each cluster. The record includes the individual measurements, the cluster ID, and the observer. The measurement record is the primary data file for the analysis. The analysis record is the record of the statistical analysis. The record includes the intra-cluster correlation estimate, the design effect, the number of clusters, the cluster size, and the analysis method. The analysis record is used to write the methods section of the final report. ### The Cluster Sampling Troubleshooting Method The troubleshooting method for cluster sampling is a structured approach to identifying and correcting problems that arise during the field season. The method is based on a set of diagnostic checks that the researcher performs at regular intervals, typically after every five clusters are sampled. The first diagnostic check is the cluster size check. The researcher compares the number of individuals measured in each cluster to the planned cluster size. If the cluster size is consistently below the planned size, the researcher must decide whether the population is smaller than expected or whether the measurement method is missing individuals. The researcher should review the field protocol and the pilot data to determine the cause. The second diagnostic check is the cluster variance check. The researcher calculates the variance of the response variable within each cluster and compares the variance across clusters. If the variance is much higher in some clusters than in others, the researcher must determine whether the clusters are different in habitat quality or whether the measurement method is inconsistent. The researcher should review the field notes for the high-variance clusters. The third diagnostic check is the cluster coverage check. The researcher maps the selected clusters and compares the map to the study area. If the clusters are concentrated in one part of the study area, the researcher must determine whether the random selection produced the concentration or whether the cluster list was incomplete. The researcher should review the cluster list and the selection record. The fourth diagnostic check is the cluster access check. The researcher records the time spent traveling to each cluster and the time spent measuring within each cluster. If the travel time is much higher than planned, the researcher must decide whether to reduce the number of clusters or to accept the higher cost. The researcher should update the cost ratio and the budget plan. The fifth diagnostic check is the cluster stability check. The researcher records any changes to the cluster boundaries or the cluster condition during the field season. If the cluster boundaries have changed, the researcher must document the change and must decide whether the cluster is still valid for the study. The troubleshooting method is recorded in the field notebook and is reviewed at the end of the field season. The researcher uses the diagnostic checks to identify the problems and to document the solutions. The solutions are reported in the final report so that the reader can assess the quality of the data. ### The Cluster Sampling Decision Log The decision log is a structured record of the decisions that the researcher makes during the study. The log includes the date, the decision, the reason for the decision, and the person who made the decision. The log is used to document the decisions that affect the sampling design and the data quality. The decision log is important because the researcher will be asked to justify the decisions in the final report and in the peer review. The log provides the evidence that the decisions were made in a transparent and reproducible way. The decision log should include the following decisions: - The decision to use cluster sampling instead of another design - The decision to define the cluster boundaries - The decision to select the number of clusters - The decision to select the cluster size - The decision to replace a lost cluster - The decision to adjust the cluster size during the field season - The decision to change the analysis method The decision log is reviewed at the end of the field season and is used to write the methods section of the final report. The log is also used to identify the decisions that were made under uncertainty and to assess the effect of the uncertainty on the results. ### The Cluster Sampling Reporting Checklist The reporting checklist for cluster sampling is a list of the items that the researcher must report in the final report. The checklist is based on the reporting guidelines for the study type and is used to ensure that the report is complete and transparent. The checklist includes the following items: - The study area and the cluster boundaries - The cluster list and the selection method - The number of clusters and the cluster size - The intra-cluster correlation and the design effect - The cost ratio and the cost plan - The cluster replacement and the cluster size adjustments - The diagnostic checks and the audit log - The decision log and the decisions that were made The researcher should use the checklist to review the final report before submission. The checklist is also used to the peer review and to the reproducibility of the study. The researcher should also consider the reporting guidelines for the study type. The reporting guidelines are available from the EQUATOR Network and are selected based on the study design and the research question. The researcher should use the reporting guideline to structure the final report and to ensure that the report is complete. The researcher should also consider the data management and sharing policy of the funding agency. The policy may require the researcher to share the data and the analysis code. The researcher should plan for the data sharing before the field season begins and should record the data in a format that can be shared. The researcher should also consider the publication ethics of the study. The researcher should follow the core practices of the Committee on Publication Ethics, which include the authorship, the peer review, the data, the conflicts of interest, and the misconduct. The researcher should record the authorship and the contributions of each team member in the decision log. The researcher should also consider the researcher identity and the record of the research. The researcher should use an ORCID identifier to link the research to the researcher and to maintain the record of the research. The researcher should record the ORCID in the decision log and in the final report. The researcher should also consider the funding and the grant policy of the funding agency. The researcher should record the grant number and the funding source in the decision log and in the final report. The researcher should follow the grant policy for the reporting and the data sharing. The researcher should also consider the data management and sharing policy of the funding agency. The researcher should plan for the data management and the data sharing before the field season begins. The researcher should record the data management plan in the decision log and in the final report. The researcher should also consider the research methods and the resources. The researcher should use the research methods resources to the research design and the analysis. The researcher should record the research methods and the resources in the decision log and in the final report. The researcher should also consider the reporting and the transparency of the research. The researcher should use the reporting guidelines and the transparency of the research. The researcher should record the reporting and the transparency in the decision log and in the final report. The researcher should also consider the ethics and the welfare of the research. The researcher should follow the ethics and the welfare of the research. The researcher should record the ethics and the welfare in the decision log and in the final report. The researcher should also consider the professional and the escalation of the research. The researcher should follow the professional and the escalation of the research. The researcher should record the professional and the escalation in the decision log and in the final report. The researcher should also consider the quality and the control of the research. The researcher should follow the quality and the control of the research. The researcher should record the quality and the control in the decision log and in the final report. The researcher should also consider the records and the measurements of the research. The researcher should follow the records and the measurements of the research. The researcher should record the records and the measurements in the decision log and in the final report. The researcher should also consider the common and the failure of the research. The researcher should follow the common and the failure of the research. The researcher should record the common and the failure in the decision log and in the final report. The researcher should also consider the professional and the escalation of the research. The researcher should follow the professional and the escalation of the research. The researcher should record the professional and the escalation in the decision log and in the final report. The researcher should also consider the frequently and the asked of the research. The researcher should follow the frequently and the asked of the research. The researcher should record the frequently and the asked in the decision log and in the final report. The researcher should also consider the related and the farming of the research. The researcher should follow the related and the farming of the research. The researcher should record the related and the farming in the decision log and in the final report. The researcher should also consider the references and the research of the research. The researcher should follow the references and the research of the research. The researcher should record the references and the research in the decision log and in the final report. The researcher should also consider the educational and the notice of the research. The researcher should follow the educational and the notice of the research. The researcher should record the educational and the notice in the decision log and in the final report. The researcher should also consider the cluster and the sampling of the research. The researcher should follow the cluster and the sampling of the research. The researcher should record the cluster and the sampling in the decision log and in the final report. The researcher should also consider the field and the biology of the research. The researcher should follow the field and the biology of the research. The researcher should record the field and the biology in the decision log and in the final report. The researcher should also consider the cluster and the selection of the research. The researcher should follow the cluster and the selection of the research. The researcher should record the cluster and the selection in the decision log and in the final report. The researcher should also consider the intra and the cluster of the research. The researcher should follow the intra and the cluster of the research. The researcher should record the intra and the cluster in the decision log and in the final report. The researcher should also consider the design and the effect of the research. The researcher should follow the design and the effect of the research. The researcher should record the design and the effect in the decision log and in the final report. The researcher should also consider the number and the cluster of the research. The researcher should follow the number and the cluster of the research. The researcher should record the number and the cluster in the decision log and in the final report. The researcher should also consider the cluster and the size of the research. The researcher should follow the cluster and the size of the research. The researcher should record the cluster and the size in the decision log and in the final report. The researcher should also consider the cost and the ratio of the research. The researcher should follow the cost and the ratio of the research. The researcher should record the cost and the ratio in the decision log and in the final report. The researcher should also consider the cluster and the visit of the research. The researcher should follow the cluster and the visit of the research. The researcher should record the cluster and the visit in the decision log and in the final report. The researcher should also consider the cluster and the audit of the research. The researcher should follow the cluster and the audit of the research. The researcher should record the cluster and the audit in the decision log and in the final report. The researcher should also consider the cluster and the decision of the research. The researcher should follow the cluster and the decision of the research. The researcher should record the cluster and the decision in the decision log and in the final report. The researcher should also consider the cluster and the field of the research. The researcher should follow the cluster and the field of the research. The researcher should record the cluster and the field in the decision log and in the final report. The researcher should also consider the cluster and the checklist of the research. The researcher should follow the cluster and the checklist of the research. The researcher should record the cluster and the checklist in the decision log and in the final report. The researcher should also consider the cluster and the sampling of the research. The researcher should follow the cluster and the sampling of the research. The researcher should record the cluster and the sampling in the decision log and in the final report. The researcher should also consider the cluster and the biology of the research. The researcher should follow the cluster and the biology of the research. The researcher should record the cluster and the biology in the decision log and in the final report. The researcher should also consider the cluster and the selection of the research. The researcher should follow the cluster and the selection of the research. The researcher should record the cluster and the selection in the decision log and in the final report. The researcher should also consider the intra and the cluster of the research. The researcher should follow the intra and the cluster of the research. The researcher should record the intra and the cluster in the decision log and in the final report. The researcher should also consider the design and the effect of the research. The researcher should follow the design and the effect of the research. The researcher should record the design and the effect in the decision log and in the final report. The researcher should also consider the number and the cluster of the research. The researcher should follow the number and the cluster of the research. The researcher should record the number and the cluster in the decision log and in the final report. The researcher should also consider the cluster and the size of the research. The researcher should follow the cluster and the size of the research. The researcher should record the cluster and the size in the decision log and in the final report. The researcher should also consider the cost and the ratio of the research. The researcher should follow the cost and the ratio of the research. The researcher should record the cost and the ratio in the decision log and in the final report. The researcher should also consider the cluster and the visit of the research. The researcher should follow the cluster and the visit of the research. The researcher should record the cluster and the visit in the decision log and in the final report. The researcher should also consider the cluster and the audit of the research. The researcher should follow the cluster and the audit of the research. The researcher should record the cluster and the audit in the decision log and in the final report. The researcher should also consider the cluster and the decision of the research. The researcher should follow the cluster and the decision of the research. The researcher should record the cluster and the decision in the decision log and in the final report. The researcher should also consider the cluster and the field of the research. The researcher should follow the cluster and the field of the research. The researcher should record the cluster and the field in the decision log and in the final report. The researcher should also consider the cluster and the checklist of the research. The researcher should follow the cluster and the checklist of the research. The researcher should record the cluster and the checklist in the decision log and in the final report. ## Frequently Asked Questions ### What is the difference between cluster sampling and stratified sampling? Cluster sampling selects a subset of groups and measures all or a subsample of individuals within the selected groups. Stratified sampling divides the population into subgroups and measures individuals from each subgroup. Cluster sampling reduces cost by visiting fewer groups. Stratified sampling reduces variance by ensuring that each subgroup is represented. ### How do I estimate the intra-cluster correlation coefficient? The intra-cluster correlation coefficient is estimated from prior data from the same species and habitat or from a pilot study. The estimate is the ratio of the between-cluster variance to the total variance. If no prior data exist, the researcher should use a range of values to test the sensitivity of the sample size calculation. ### What is the design effect in cluster sampling? The design effect is the ratio of the variance under cluster sampling to the variance under simple random sampling. It is calculated from the intra-cluster correlation and the cluster size. The design effect is used to adjust the sample size calculation. ### How many clusters are needed for a field study? The number of clusters is calculated from the desired precision, the variance, the intra-cluster correlation, and the cluster size. The researcher should calculate the number of clusters and should select more clusters than the minimum to allow for field losses. ### Can I use cluster sampling for a small study area? Cluster sampling is most useful for large study areas where the cost of reaching an individual is high. For a small study area, simple random sampling may be more efficient because the cost of reaching an individual is low and the statistical efficiency is higher. ### How do I analyze data from a cluster sampling design? The data must be analyzed with a method that accounts for the cluster. The researcher should use a mixed-effects model with the cluster as a random effect or a generalized estimating equation. The analysis must include the cluster identifier as a variable. ### What are the common mistakes in cluster sampling? The common mistakes are ignoring the intra-cluster correlation, selecting clusters that are too similar, using too few clusters, and changing the cluster boundaries during the study. The researcher should avoid these mistakes by following the workflow and the quality controls. ### How do I report the results of a cluster sampling study? The researcher should report the number of clusters, the cluster size, the intra-cluster correlation, and the design effect. The researcher should also report the method of cluster selection and the analysis method. The reporting should follow the relevant reporting guideline for the study type. ## Using the Evidence | Source | Best use in this topic | Important limitation | |---|---|---| | [Research Methods Resources](https://www.ncbi.nlm.nih.gov/books) | official guidance | Check the linked page for current local requirements | | [EQUATOR Network](https://www.equator-network.org/) | official guidance | Check the linked page for current local requirements | | [Core Practices](https://publicationethics.org/core-practices) | official guidance | Check the linked page for current local requirements | ## Related Diagnostic Guides - [Animal Biology](/blog/guides/animal-biology) - [Animal Research](/blog/guides/animal-research) - [Biology Facts](/blog/guides/biology-facts) ## References and Further Reading - [Research Methods Resources](https://www.ncbi.nlm.nih.gov/books). National Library of Medicine. - [EQUATOR Network](https://www.equator-network.org/). EQUATOR Network. - [Core Practices](https://publicationethics.org/core-practices). Committee on Publication Ethics. - [NIH Grants and Funding](https://grants.nih.gov/). National Institutes of Health. - [ORCID for Researchers](https://info.orcid.org/researchers). ORCID. - [Data Management and Sharing Policy](https://sharing.nih.gov/data-management-and-sharing-policy). National Institutes of Health. - [Laboratory Biosafety Manual](https://www.who.int/publications/i/item/9789240011311). World Health Organization. - [NCBI Literature Resources](https://www.ncbi.nlm.nih.gov/). National Center for Biotechnology Information. - [Single-Cell RNA Sequencing Analysis: A Step-by-Step Overview.](https://pubmed.ncbi.nlm.nih.gov/33835452). Methods in molecular biology (Clifton, N.J.), 2021. - [Does Sequence Clustering Confound AlphaFold2?](https://pubmed.ncbi.nlm.nih.gov/40780395). Journal of molecular biology, 2025. - [Bayesian clustering with uncertain data.](https://pubmed.ncbi.nlm.nih.gov/39226325). PLoS computational biology, 2024. > This article is educational and does not replace validated laboratory procedures, institutional biosafety review, manufacturer instructions, or professional interpretation.