METHOD GUIDE
Reliability Analysis in SPSS
Reverse-Coded Items, Item-Total Tables and Omega
A reliability test in SPSS can return an alpha that looks acceptable for the wrong reasons, or one that turns negative because of a single unreversed item. Item coding comes first, the item-total columns show which items carry the coefficient, and factor evidence decides whether omega belongs beside it.
IN SHORT
What Alpha Summarizes and What It Leaves Open
A reliability analysis estimates how consistently the items of a multi-item scale measure the same attribute, and alpha is the coefficient SPSS reports by default. It applies to items scored in one direction on a shared response format, and it describes the scores of the sample at hand, not the instrument for every use. To run a reliability test in SPSS, open Analyze, Scale, Reliability Analysis, enter the items, keep the Alpha model, and request the item and scale descriptives, Scale if item deleted and the inter-item correlations under Statistics (labels vary by version). OnlineSPSS, a specialist statistical consultancy for academic, clinical and professional research, carries out and reviews this analysis through an in-house PhD-level statistician.
ANALYSIS SUPPORT
Analysis Support for This Method
Once a scale's coefficient is accepted, its rating items and summed scores still need an analysis that suits them, the focus of Likert Scale Analysis.
Reliability Analysis (Cronbach's Alpha)
The reliability analysis fitted within your study design, with the item coding, the dimensionality question, the choice of coefficient and the reporting agreed for your scale.
Exploratory Factor Analysis (EFA)
The check on how many dimensions the items form, for when a scale has not been validated before and alpha cannot say whether one attribute underlies it.
BEFORE YOU RUN IT
Getting the Items Ready for Alpha
A single coefficient hides the work done on the items, so that preparation is decided and written down before any coefficient is computed.
Reverse-Coded Items Negatively worded items are recoded so that a high score always means more of the attribute. SPSS does not reverse them for you, and one unreversed item pulls alpha down or turns it negative.
One Attribute per Alpha Alpha is computed for a set of items meant to measure one attribute. Pooling items from several subscales gives a number that rises with the item count and says little about any one of them.
Enough Items, Common Format Items share a response format, or are standardized, and there are enough of them for a stable estimate. With very few items alpha is unstable, and a two-item measure is better reported with the Spearman-Brown coefficient.
Missing Responses The Reliability procedure uses only cases complete on every item entered, so scattered gaps shrink the sample. The N in the Case Processing Summary is compared with the intended sample.
IN SPSS
How to Run a Reliability Test in SPSS
In IBM SPSS the recoding happens under Transform before the Reliability dialog is opened, so the steps start there. Dialog labels may differ in your version.
- Reverse-code negatively worded items into new variables, and check the new codes against the old.
- Choose Analyze, Scale, Reliability Analysis and move the items of one scale into the Items box.
- Keep Alpha as the Model.
- Under Statistics, tick Item, Scale and Scale if item deleted under Descriptives for, and Correlations under Inter-Item.
- Run the analysis and read the Case Processing Summary and Reliability Statistics.
- Read the Item-Total Statistics, then repeat the analysis separately for each subscale.
READING THE OUTPUT
Item-Total Statistics and the Other Tables Behind One Alpha
The Reliability Statistics table gives one number, while item-total statistics in SPSS show which items produce it and which pull it down.
| Output element | What it tells you | What to check |
|---|---|---|
| Case Processing Summary | The valid, excluded and total cases, where a gap on any item removes the case | That the excluded count is explained and the N matches the intended sample |
| Reliability Statistics | Cronbach’s Alpha, Cronbach’s Alpha Based on Standardized Items and N of Items | That a negative alpha is investigated and not reported, and the unstandardized value is used unless items were standardized |
| Item Statistics | The mean, standard deviation and N of each item | That an item with almost no spread, or a ceiling, is noticed because it limits its correlation with the rest |
| Inter-Item Correlation Matrix | The correlation between every pair of items | That negative correlations point to unreversed items, and near-identical items to redundant wording |
| Item-Total Statistics: Corrected Item-Total Correlation | Each item’s correlation with the sum of the other items | That values near zero or negative are traced to wording or coding before any item is removed |
| Item-Total Statistics: Cronbach’s Alpha if Item Deleted | The alpha the scale would have without that item | That a small rise is not a reason to drop an item whose content the attribute requires |
| Scale Statistics | The mean, variance, standard deviation and N of Items for the total score | That these describe the summed score that is analyzed later, with its direction stated |
REPORTING
Reporting Alpha With the Scale, the Items and Their Direction
Reliability belongs in the measurement section, written for each scale or subscale with its items and their direction.
Describe the scale, the number of items, the response format and which items were reverse-coded, then report Cronbach’s α = ___ with its 95% CI [___, ___], N = ___, and the range of corrected item-total correlations. The Reliability Statistics table prints alpha without an interval, but the Intraclass correlation coefficient option (Two-Way Mixed, Consistency) prints one, because its Average Measures value equals alpha. Alpha equals reliability only when the items load equally on one dimension, and with unequal loadings it can misstate it, so omega, which allows unequal loadings, is reported beside it where factor evidence points that way. Recent SPSS versions list Omega as a Model in the Reliability dialog; where yours does not, it is computed in another program. Present reliability as a property of the scores in this sample, and report it again for each sample and each subscale.
COMMON MISTAKES
Alpha Reporting Habits That Weaken a Measurement Section
The coefficient is easy to produce, so reviewers look at what stands behind it: the items, the dimensionality and the thresholds.
Unreversed Items A negative alpha, with a warning about a negative average covariance, often means a coding problem. The codes are fixed first, and a negative value is never reported as the scale's reliability.
Alpha Read as One Dimension A high alpha can arise from many items that measure two or three related attributes. Evidence from factor analysis, not the size of alpha, shows whether one dimension underlies the items.
Items Deleted to Raise Alpha Removing items one at a time to push alpha up fits the scale to the sample and can cut content the construct needs. Deletion is argued from item content and reported.
A Fixed Pass Mark Quoting .70 as a universal standard ignores the purpose, the number of items and the stakes of the decision. The value is read against how the scores will be used and the literature on that scale.
Consistency Taken as Validity A high alpha shows that the items vary together in this sample, not that they measure the intended construct. Evidence of validity comes from the item content, the factor structure and relations with other measures.
Reliability Carried Over Citing the alpha from the instrument's original study does not establish reliability in your sample. It depends on the sample, so it is computed on your data and reported.
Alpha already computed? Independent Statistical Review checks the item coding, the item-level tables and the reliability statement before you submit.
FREQUENTLY ASKED QUESTIONS
Questions About Reliability Output in SPSS
When a reliability table looks wrong, the cause usually sits in one column or one item, and these answers point to where.
Why is my Cronbach's alpha in SPSS negative?
A negative alpha means the average covariance among the items is negative, and SPSS prints a warning that this violates the reliability model. The most frequent cause is a negatively worded item that was never reverse-coded. The Inter-Item Correlation Matrix and the Corrected Item-Total Correlation column show which item runs against the others. After recoding, the value is usually positive, and if it stays negative the items do not share a common attribute.
Which column of the item-total table shows that an item is weak?
The Corrected Item-Total Correlation column, because it correlates each item with the sum of the other items. Values near zero or negative mark an item that does not move with the rest, and wording, coding or content is checked before anything is removed. The Cronbach’s Alpha if Item Deleted column is read afterward, since a small gain in alpha does not justify dropping an item the construct needs.
Does SPSS give a confidence interval for alpha?
Not in the Reliability Statistics table, but the Intraclass correlation coefficient option under Statistics supplies one: with the Two-Way Mixed model and Consistency type, the Average Measures row equals alpha and carries a 95% confidence interval. Reporting that interval shows how uncertain alpha is when the sample or the number of items is small, and a bootstrap or another program gives the same information where preferred.
Can OnlineSPSS review my item coding and reliability statement?
Yes. An in-house PhD-level statistician can check the reverse-coded items, the item-level tables and the coefficient you report, advise when omega should accompany alpha, or carry out the analysis on your data. Where the coefficients are final, Results Reporting & Interpretation sets out the measurement table and its description. The measurement decisions and the conclusions remain yours.
How do you protect my research data and confidentiality?
Your data and project details are handled confidentially, and project information is not published. Researcher names and institutions are not used publicly without permission. An NDA is available where required and can be arranged before sensitive materials are shared. Read more about confidentiality and data security.
What happens when I request an assessment, and how long does it take?
You share your research question, design and relevant materials, such as the dataset or the output you have, and a statistically trained Research Coordinator reviews the request and responds the same day. The analysis, deliverables, timeline and one fixed project fee are then confirmed before any work begins, so the timeline follows from the agreed scope.
Still have a question? Find more answers in our full FAQ.
NEXT STEP
Does Your Alpha Describe
the Scale You Meant?
List the items with their response format and which were reverse-coded, and add the reliability tables. The scope confirmed after review states which coefficients will be examined.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014