Partially Balanced Incomplete Block Design (PBIBD)
Analyze incomplete-block experiments when a fully balanced BIBD cannot be constructed. A PBIBD groups treatment pairs into associate classes, each with its own concurrence, keeping the design practical while retaining structure.
Run PBIBD Analysis →What is a Partially Balanced Incomplete Block Design?
A Partially Balanced Incomplete Block Design (PBIBD) is an incomplete-block design used when the strict requirements of a Balanced Incomplete Block Design cannot be met. Rather than forcing every pair of treatments to appear together the same number of times, a PBIBD allows the pairwise concurrence to take a small number of distinct values.
Treatment pairs are sorted into associate classes. Two treatments that are first associates appear together λ1 times; second associates appear together λ2 times, and so on. The associate structure is defined by an association scheme, such as a group-divisible or triangular scheme.
This partial balance makes designs feasible for many treatment and block-size combinations where no BIBD exists, at the cost that comparisons between first associates and between second associates are estimated with slightly different precision.
In plain terms: A full BIBD isn't always possible for your numbers. A PBIBD bends the rule: instead of every pair meeting equally often, treatments are sorted into groups, and pairs within a group meet one number of times while pairs across groups meet another. It's a practical compromise that keeps most of the balance.
Design Concepts
Associate Classes
Treatment pairs are grouped by association. First associates share concurrence λ1; second associates share λ2. Most PBIBDs use two associate classes.
Association Schemes
Common schemes include group-divisible, triangular and Latin-square-type. The scheme fixes how many associates of each class every treatment has.
Partial Balance
Comparisons within an associate class share precision, but the two classes differ. This is the relaxation that makes the design constructible.
Key Formulas
Reading a PBIBD Analysis
As with a BIBD, treatment effects are block-adjusted, so raw averages are not compared directly. The adjustment uses the associate structure.
Because precision differs by associate class, the standard error of a difference depends on whether the two treatments are first or second associates. Report and compare using the correct standard error for each pair.
Assumptions & Validation
Valid Association Scheme
A recognized association scheme (group-divisible, triangular, etc.) must fit your treatment count and block size.
If violated: If none fits, reconsider treatment numbering or block size.
Additivity
Block and treatment effects add with no interaction.
If violated: Interaction requires a different design.
Normality
Residuals are approximately normal.
If violated: Use a rank-based method or transform.
Equal Variance
Error variance is constant across blocks.
If violated: Apply a variance-stabilizing transformation.
⚠️ Check assumptions first
A PBIBD is only meaningful within a defined association scheme; the analysis and the standard errors depend on which treatments are first versus second associates. Do not treat a PBIBD as if it were fully balanced: comparisons across associate classes have different precision, and reporting a single standard error for all pairs is incorrect.
When NOT to Use Partially Balanced Incomplete Block Design (PBIBD)
A BIBD Exists
If valid BIBD parameters are available, prefer the fully balanced design, since all comparisons then share equal precision.
Complete Blocks Possible
If blocks can hold every treatment, a Randomized Block Design is simpler and fully efficient.
No Sensible Association Scheme
If no association scheme fits your treatment structure, an incomplete-block design may not be appropriate; reconsider the layout.
Industry Applications
Large Variety Trials
Plant-breeding programs with many varieties and small field blocks, where a BIBD is impossible but partial balance is achievable.
Industrial Screening
Comparing many process variants across limited test slots when full balance cannot be scheduled.
Sensory Studies
Panels with many products and short evaluation sessions, grouped by an association scheme.
Genetics & Trials
Structured comparisons where treatments naturally fall into groups matching an association scheme.
Frequently Asked Questions
How does a PBIBD differ from a BIBD?
In a BIBD every pair of treatments appears together the same number of times, giving equal precision for all comparisons. A PBIBD allows the pairwise concurrence to take two or more values depending on an association scheme, so treatments are grouped into associate classes. This relaxation makes designs constructible when no BIBD exists, at the cost of unequal precision across classes.
What is an associate class?
An associate class groups pairs of treatments by how often they occur together. Typically two treatments are either first associates, appearing together lambda-one times, or second associates, appearing together lambda-two times. The pattern of who is a first or second associate of whom is fixed by the association scheme.
What is an association scheme?
An association scheme is the rule that defines the associate relationships among treatments. Common schemes include group-divisible, where treatments are split into groups; triangular; and Latin-square-type. The scheme determines how many first and second associates each treatment has and underlies the whole analysis.
When should I choose a PBIBD?
Choose a PBIBD when you need an incomplete-block design, because blocks cannot hold all treatments, but no valid BIBD exists for your combination of treatment count and block size. The PBIBD keeps most of the structure and efficiency of a balanced design while remaining feasible to construct.
Are all treatment comparisons equally precise in a PBIBD?
No. Comparisons between first associates share one level of precision and comparisons between second associates share another. This is the defining trade-off of partial balance. When reporting results you must use the standard error appropriate to the associate class of the two treatments being compared.
Do I still adjust treatment effects for blocks?
Yes. As in any incomplete-block design, no treatment appears in every block, so raw treatment averages are biased. Treatment effects are adjusted for blocks using intra-block information and the associate structure, and only these adjusted effects should be compared.
Analyze Incomplete-Block Experiments When BIBD Fails
Get associate-class-aware, block-adjusted comparisons from your PBIBD. Free during Beta.
Run PBIBD Analysis →