Written and maintained by CASRAI Editorial Board
Published
On this page
SPSS has no critical-value table or menu item. To get a critical value in SPSS Statistics, open Transform > Compute Variable and evaluate an inverse distribution function: IDF.NORMAL for a z value, IDF.T for a t value, IDF.CHISQ for chi-square and IDF.F for F. Each function takes a cumulative (left-tail) probability, so a two-tailed test at α = 0.05 uses 0.975, not 0.05.
What critical value in SPSS is used for in research
A critical value is the cutoff a test statistic must pass for a result to be significant at a chosen α. SPSS output usually skips that step and prints an exact p-value in the “Sig.” column, which you compare with α directly (see the guide to reading SPSS output below). You still need the critical value itself in several situations: checking a hand calculation or a published table, writing a methods section that states the rejection region, planning a study where the cutoff feeds a power or sample-size argument, or teaching how a test decision is made. For the concept and for the printed-table route (including Excel, R and Python), see the companion guide on how to find a critical value; this page covers only the SPSS route.
SPSS calls the needed functions “inverse distribution functions”. The IBM Command Syntax Reference defines IDF.T(prob, df) as returning “the value from Student’s t distribution, with specified degrees of freedom df, for which the cumulative probability is prob” (IBM SPSS Statistics 24 Command Syntax Reference, Universals chapter, read 2026-11-17). That is exactly what a critical value is: the point of the distribution that cuts off a stated cumulative probability.
How to access the Compute Variable tool in SPSS
There is nothing extra to install. The inverse distribution functions are part of the SPSS Statistics base system and are available wherever you can run Transform > Compute Variable. SPSS Statistics is an IBM product; most researchers reach it through an institutional licence or a licence obtained directly from IBM. The Compute Variable dialog and the IDF functions are documented in IBM’s own Computing Variables page and in the Command Syntax Reference, which are the sources used for every syntax statement below.
One practical requirement: Compute writes its result into a variable, so the Data Editor needs at least one case. Open any dataset you are working on, or type a single placeholder value into the first cell of a new one, as the George Mason University SPSS tutorial does by typing 0.975 into the data editor before computing.
How to compute a critical value in SPSS step by step
The steps below follow IBM’s Computing Variables page (Transform > Compute Variable, Target Variable box, Expression field, function groups) and the IDF definitions in the syntax reference.
- Decide the distribution (z, t, chi-square or F), α, and whether the test is one-tailed or two-tailed.
- Convert α to the cumulative probability SPSS needs. For a right-tail or two-tailed critical value, the probability is 1 − α (one-tailed) or 1 − α/2 (two-tailed, upper cutoff). For a left-tail cutoff, the probability is α itself.
- Choose Transform > Compute Variable….
- In the Target Variable box, type a new variable name such as
crit. - In the Expression field, type the function with your numbers, for example
IDF.T(0.975, 7). Alternatively select the All function group, double-click the function to paste it, and replace the question marks with your parameters, as IBM describes. - Click OK. The value is written to the new variable in every case. Read it from the Data Editor.
- If the value shows too few digits, open Variable View and increase the Decimals setting for the variable; the display default is two decimals, so more digits are hidden, not lost.
The syntax-file version
The same operation can be pasted or typed as syntax. Using the function names the reference documents:
COMPUTE crit_z = IDF.NORMAL(0.975, 0, 1). COMPUTE crit_t = IDF.T(0.975, 7). COMPUTE crit_chi = IDF.CHISQ(0.95, 3). COMPUTE crit_f = IDF.F(0.95, 3, 100). EXECUTE.
Each COMPUTE creates one variable holding the same number on every row, and EXECUTE runs the pending transformations. To build a whole critical-value table in one pass, put the degrees of freedom in a variable (say df, values 1 to 30) and write COMPUTE crit_t = IDF.T(0.975, df).; SPSS then returns one critical value per row.
Which probability goes in: the one trap
The function documentation says prob is a cumulative probability. IBM’s own example for chi-square is that the value significant at the 0.05 level with 3 degrees of freedom is IDF.CHISQ(0.95,3), and for F that the value significant at 0.05 with 3 and 100 degrees of freedom is IDF.F(0.95,3,100). Typing 0.05 there returns the lower 5% point of the distribution, which is a valid number but not the critical value for a right-tailed test. For a two-tailed t test at α = 0.05 you need the point with 97.5% of the distribution below it, so the probability is 0.975.
The data table below maps each distribution to its SPSS function, the probability argument for a 5% test and the value it returns. We computed the returned values independently with SciPy’s percent-point functions and they match the rounded figures shown.
How to read the output
There is no output window entry: the result lives in the new variable. Three habits keep it reliable.
- Compare with the sign and tail of your test. A two-tailed t test has critical values of −crit and +crit. IDF.T(0.975, 7) returns the positive one, 2.365; the negative one is the same number with a minus sign. A computed t statistic beyond ±2.365 in a 7-degree-of-freedom test is significant at α = 0.05, two-tailed.
- Chi-square and F are right-tailed. These statistics are never negative and the rejection region is the upper tail, so use 1 − α. IDF.CHISQ(0.95, 3) returns 7.815; a chi-square statistic above that is significant at 0.05 with 3 degrees of freedom.
- Cross-check against a printed result. Run a t test, note the degrees of freedom and the t statistic in the Output Viewer, compute the critical value for those degrees of freedom, and confirm that the statistic exceeds the critical value exactly when “Sig.” is below α. If it does not, the usual cause is a one- versus two-tailed mix-up in step 2. The guide to reading SPSS output explains the tables you are checking against, and the t-distribution table guide lists printed values you can compare with yours.
For the z case, the George Mason University tutorial shows the same logic: a value of 0.975 passed to IDF.NORMAL with mean 0 and standard deviation 1 gives 1.96, the familiar two-tailed 5% cutoff. See z-scores for what that cutoff means in standard-deviation units.
Worked check
Suppose a paired-samples t test on eight pairs gives t = 2.9 with 7 degrees of freedom, and you test at α = 0.05, two-tailed. Compute IDF.T(0.975, 7), which returns 2.365. Because 2.9 exceeds 2.365, the result is significant, and the “Sig. (2-tailed)” value in the output should read below 0.05. For a one-tailed test at the same α, use IDF.T(0.95, 7), which returns 1.895.
Troubleshooting critical values in SPSS
- The value looks rounded (for example 2.37 instead of 2.3646). That is a display setting, not a calculation limit. Raise Decimals in Variable View, as the George Mason tutorial notes.
- The value looks too small or negative. You probably passed α instead of 1 − α, or the lower tail. For a right-tailed critical value, pass the cumulative probability 1 − α.
- Very large degrees of freedom. An IBM support note records an inaccuracy in IDF.T for very large degrees of freedom that was fixed in SPSS 20 Fixpack 2. If you run an older release and need large-df t values, check your version.
A note on versions: the function definitions quoted here are from the Command Syntax Reference for SPSS Statistics 24, and the Compute Variable dialog description is from IBM’s version 23 documentation. The IDF functions are long-standing members of the SPSS expression language, but check the reference for your installed release if a function name is rejected.
Free alternatives to SPSS for critical values
Any environment that exposes a quantile function gives the same numbers. The R functions below are documented in the R stats manual pages; their argument is also a cumulative probability, with a lower.tail option.
| Distribution | SPSS function | R function (stats package) |
|---|---|---|
| Normal (z) | IDF.NORMAL(prob, mean, stddev) | qnorm(p, mean = 0, sd = 1, lower.tail = TRUE) |
| Student’s t | IDF.T(prob, df) | qt(p, df, ncp, lower.tail = TRUE) |
| Chi-square | IDF.CHISQ(prob, df) | qchisq(p, df, ncp = 0, lower.tail = TRUE) |
| F | IDF.F(prob, df1, df2) | qf(p, df1, df2, ncp, lower.tail = TRUE) |
In R, qt(0.975, 7) returns the same 2.365 as the SPSS expression, and setting lower.tail = FALSE lets you pass α directly instead of 1 − α. For running the tests that produce the statistics you compare against, see analysing data in SPSS.
Sources
- IBM SPSS Statistics 24 Command Syntax Reference (Universals: inverse distribution functions), IBM, read 2026-11-17.
- Computing Variables, IBM SPSS Statistics 23 documentation, IBM, read 2026-11-17.
- SPSS worked example: critical values with IDF functions, George Mason University, read 2026-11-17.
- The Student t Distribution (qt), R stats manual, read 2026-11-17.
- The (non-central) Chi-Squared Distribution (qchisq), R stats manual, read 2026-11-17.
- The F Distribution (qf), R stats manual, read 2026-11-17.
- The Normal Distribution (qnorm), R stats manual, read 2026-11-17.
Data: SPSS inverse distribution functions for 5% critical values
Maps each distribution to the SPSS IDF function, the cumulative probability argument for a 5% test (one- or two-tailed) and the critical value returned. Syntax and the cumulative-probability convention come from the IBM SPSS Statistics Command Syntax Reference; returned values were recomputed independently with SciPy percent-point functions.
| Case | SPSS expression | Cumulative probability argument | Critical value |
|---|---|---|---|
| z, two-tailed, alpha 0.05 | IDF.NORMAL(0.975, 0, 1) | 0.975 | 1.96 |
| z, one-tailed, alpha 0.05 | IDF.NORMAL(0.95, 0, 1) | 0.95 | 1.645 |
| t, df 7, two-tailed, alpha 0.05 | IDF.T(0.975, 7) | 0.975 | 2.365 |
| t, df 7, one-tailed, alpha 0.05 | IDF.T(0.95, 7) | 0.95 | 1.895 |
| t, df 20, two-tailed, alpha 0.05 | IDF.T(0.975, 20) | 0.975 | 2.086 |
| Chi-square, df 3, alpha 0.05 | IDF.CHISQ(0.95, 3) | 0.95 | 7.815 |
| Chi-square, df 1, alpha 0.05 | IDF.CHISQ(0.95, 1) | 0.95 | 3.841 |
| F, df 3 and 100, alpha 0.05 | IDF.F(0.95, 3, 100) | 0.95 | 2.696 |
| F, df 2 and 30, alpha 0.05 | IDF.F(0.95, 2, 30) | 0.95 | 3.316 |
Sources
- IBM SPSS Statistics 24 Command Syntax Reference (accessed 2026-11-17)
- Computing Variables, IBM SPSS Statistics 23 (accessed 2026-11-17)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







