# SPSS syntax to produce two-tailed test for regression

I have some syntax for SPSS and it produces everything that I need perfectly, although for some reason it spits it out one tailed instead of two tailed. Is there any way of editing this syntax so that it uses a two tailed analyses?

REGRESSION
/DESCRIPTIVES MEAN STDDEV CORR SIG N
/MISSING LISTWISE
/STATISTICS COEFF OUTS CI(95) R ANOVA CHANGE ZPP
/CRITERIA=PIN(.05) POUT(.10)
/NOORIGIN
/DEPENDENT SumF1
/METHOD=ENTER Gender
/METHOD=ENTER Age
/METHOD=ENTER Student BlueCollar WhiteCollar
/METHOD=ENTER Single Married DeFacto
/SCATTERPLOT=(*ZPRED ,*ZRESID)
/RESIDUALS HISTOGRAM(ZRESID) NORMPROB(ZRESID).

• It would be helpful if you describe your variables, what do they mean, and also name the output table in which you find one-tailed tests instead of two-tailed what you expected. Commented Nov 22, 2011 at 13:01

Surprisingly you can not specify this for the output. Are you sure yours is displaying the one-tailed test? The example below (and this is using version 19) is displaying the two-tailed test as the default for me.

data list free / v1 v2.
begin data
1 10
2 12
3 14
4 19
end data.

REGRESSION
/MISSING LISTWISE
/STATISTICS COEFF OUTS R ANOVA
/CRITERIA=PIN(.05) POUT(.10)
/NOORIGIN
/DEPENDENT v1
/METHOD=ENTER v2.


This produces the table of coefficient estimates below;

So I have two degrees of freedom for the t-test (since I estimated two parameters, v2 and the intercept), and an estimate of 5.581. Looking at the t-test table on wikipedia you can see a t-value of 5.581 for a two-tailed test would produce a p-value somewhere in between .02 and .05 (that table is slightly odd compared to most t-tables, and 100 minus the percentage produces the p-value of interest). You should also notice from that table that the p-value's associated with one-tailed tests are simply the p-value of the two-tailed test divided by two. So even though SPSS does not give the option to change whether you use a one tailed or two tailed test, it is fairly trivial to do do the calculation yourself. (A more exact two-tailed p-value from this online calculator suggests it is .0306, which is rounded up in the table below)

So if you wanted to go from the two-tailed test to the one-tailed test, you would just divide the p-value by 2. Conversely if you wanted to go from a one-tailed test to a two-tailed test, you would just multiple the p-value by 2.

There are no one-tailed tests on coefficients from Regression except, of course, that an F test is always one tailed.