Hypothesis Direction and Sidedness

Before running a confirmatory analysis, you should decide whether your hypothesis is directional — whether you are predicting that one group will be higher, lower, or simply different from another.

This decision affects both the statistical test and how the result is interpreted. Licklider asks you to make it explicit when it matters.

Two-sided vs one-sided tests

Two-sided (non-directional) Tests whether the groups differ in either direction. The p-value reflects the probability of observing a difference as large as the one found, in either direction, under the null hypothesis.

Use two-sided tests when:

• You have no prior basis for predicting the direction
• Your hypothesis is that the groups differ, not that one is specifically higher or lower
• You are in an exploratory phase

Two-sided testing is the default in Licklider.

One-sided (directional) Tests whether one group is specifically higher or specifically lower than another. The p-value reflects the probability of observing a difference in the predicted direction.

Use one-sided tests only when:

• The direction was pre-specified before data collection, based on prior evidence or theory
• You would not act on a significant result in the opposite direction
• The directionality is documented in a pre-registration or analysis plan

One-sided tests are more powerful when the direction is correctly specified, but they are anti-conservative and misleading when applied post-hoc to the direction that happened to show significance.

When Licklider asks about direction

Licklider does not ask about hypothesis direction for every analysis. The question arises when:

1. The analysis intent is set to confirmatory or publication-ready, and
2. The test is a two-group comparison or a significance annotation

For exploratory analyses, directionality is not required and no confirmation step appears.

When the conditions above are met, Licklider presents a confirmation step with the following options:

Two-sided The analysis tests for a difference in either direction. No directional prediction is claimed.

One-sided (with preregistered direction) The analysis tests for a difference in a specific direction. Selecting this option requires you to specify whether you are testing for an increase or a decrease, and to confirm that the direction was pre-specified.

If the directionality is not resolved before a claim-bearing export, the figure is marked as provisional.

How the declaration connects to the calculation

When you select a direction in the confirmation step, Licklider uses that declaration directly in the statistical calculation. The declared direction determines the alternative parameter passed to the test engine.

This applies to:

• Independent-samples t-test (Welch and Student's)
• Paired t-test
• Mann-Whitney U test
• Wilcoxon signed-rank test

For ANOVA and chi-square tests, directionality does not apply — these tests are inherently non-directional by design.

What you will see in practice

Directionality is not just a hidden test setting. When it matters, you should expect to see:

• A confirmation step asking whether the claim is two-sided or one-sided
• An explicit increase or decrease choice when one-sided is selected
• A provisional state for claim-bearing export if the direction has not been resolved
• A recorded rationale when a one-sided test is used
• The rationale reflected in the methods text draft

These outputs do not create a separate result on their own. Instead, they determine how the p-value is calculated, how the claim is framed, and whether the figure is eligible for claim-bearing use.

Power calculations and sidedness

Power calculations in Licklider use two-sided tests by default. One-sided power calculations are not currently supported.

If your analysis plan involves a one-sided test, note that the required sample size from the power calculator is slightly conservative — a one-sided test at the same alpha requires a smaller sample than a two-sided test. The difference is modest for typical alpha and power settings.

For more detail — see Power Analysis and Sample Size Calculation.

Recording the rationale

When a one-sided test is selected, Licklider asks for a brief rationale confirming that the direction was pre-specified. This rationale is recorded in the figure's disclosure and appears in the methods text draft.

If no rationale is provided, the figure remains in an unresolved state and is not eligible for claim-bearing export.

What Licklider can and cannot determine automatically

Licklider can ask when directionality matters, record the declared direction, and pass that declaration into the test calculation.

However, Licklider cannot determine automatically whether a one-sided hypothesis was genuinely pre-specified before data analysis. In particular, Licklider cannot reliably detect:

• Whether the directional claim was chosen before looking at the data or after seeing which direction looked favorable
• Whether a pre-registration or analysis plan truly supports the stated directional hypothesis
• Whether the researcher would genuinely treat a strong effect in the opposite direction as non-supportive
• Whether a previously generated figure still meets a newly declared directional standard without being reviewed again

These limits matter because post-hoc direction choices can make evidence look stronger than it really is. A one-sided test chosen after seeing the data can produce an anti-conservative interpretation even if the calculation itself is mechanically correct.

That is why Licklider requires an explicit declaration, records a rationale for one-sided claims, and leaves the scientific justification to documented study planning rather than trying to infer it from the data alone.

Design Rationale & References

This page follows a simple rule: the strength and direction of a confirmatory claim should be declared before the result is treated as evidence. That is why Licklider defaults to two-sided testing, restricts one-sided use to explicitly declared directional claims, and records a rationale when a one-sided test is selected.

The two-sided default is intentional. In most research settings, the safer default is to test for a difference in either direction unless a directional hypothesis was defined in advance. One-sided tests can be more powerful, but that gain is only defensible when the direction was truly pre-specified and the opposite direction would not be used to support the claim [1, 2].

The provisional state and rationale requirement are also deliberate. They are not about changing the mathematics of the test after the fact; they are about preventing a directional claim from appearing fully justified when the study plan has not made that justification explicit.

Power calculations remain two-sided by default because that is the supported planning path in Licklider today, and it is the more conservative assumption when a directional claim has not yet been firmly justified.

1. Ruxton, G. D., & Neuhäuser, M. (2010). When should we use one-tailed hypothesis testing? Methods in Ecology and Evolution, 1(2), 114-117. https://doi.org/10.1111/j.2041-210X.2010.00014.x
2. Fisher, R. A. (1935). The Design of Experiments. Oliver and Boyd.

What this page does not cover

• How analysis intent is set — see Outcome Type and Analysis Intent
• Multiple comparisons within a figure — see Multiplicity and Analysis Families
• Power calculations — see Power Analysis and Sample Size Calculation
• Effect size reporting — see Effect Size, CI, and N Reporting