Does anyone know of a method or approach in SPSS to contrast singular mean scores of overall samples, based only on the overall mean value, standard deviation figure and sample size? The data is not available, so only the overall values from that sample.
see basic statistics book under t test. quite easy with calculator. best, D Booth
Hello Nathan. I think you are asking how to carry out a t-test from summary data (using SPSS), but can't tell if you mean a one-sample t-test or an unpaired t-test. Here are examples of both. HTH.
* One-sample t-test from summary data.
* Use the example on this webpage:
* https://stats.idre.ucla.edu/spss/output/t-test/.
NEW FILE.
DATASET CLOSE ALL.
DATA LIST LIST / n (F5.0) m sd mu0 (3F8.3).
BEGIN DATA
200 52.775 9.47859 50
END DATA.
COMPUTE df = n-1.
FORMATS df (F5.0).
COMPUTE Diff = m-mu0.
COMPUTE SE = sd/SQRT(n).
COMPUTE tcrit = ABS(IDF.T(.025,df)).
COMPUTE LL = diff-SE*tcrit.
COMPUTE UL = diff+SE*tcrit.
COMPUTE tobs = (m-mu0)/SE.
COMPUTE p = 2*(1-CDF.T(ABS(tobs),df)).
FORMATS SE to p (F8.3).
LIST.
* Unpaired t-test from summary data.
* Note that t^2 = F for an ANOVA on the same data.
* Therefore, you can use ONEWAY command with MATRIX input.
* Use example from this webpage:
* https://stats.idre.ucla.edu/spss/output/t-test/.
DATA LIST LIST / ROWTYPE_ (a8) Grp (f5.0) VARNAME_ (a8) Y (f8.2) .
BEGIN DATA
"MEAN" 0 "Male" 50.1209
"STDDEV" 0 "Male" 10.30516
"N" 0 "Male" 91
"MEAN" 1 "Female" 54.9908
"STDDEV" 1 "Female" 8.13372
"N" 1 "Female" 109
END DATA.
* Perform the analysis with the ONEWAY procedure using
* MATRIX input .
ONEWAY Y BY grp /MATRIX = in(*) /STATISTICS=WELCH BROWNFORSYTHE.
COMPUTE t_eva = SQRT(13.943035).
COMPUTE t_evna = SQRT(13.369060).
FORMATS t_eva t_evna (F8.3).
LIST t_eva t_evna /CASES=1.
By the way, both of these tests can be done much more easily (IMO) using "immediate" versions of the t-test command in Stata (if you have access to it). Here's the code:
* One-sample t-test
ttesti 200 52.775 9.47859 50
* Unpaired t-test
* 1) Pooled variances version
ttesti 91 50.1209 10.30516 109 54.9908 8.13372
* 2) Welch's t-test
ttesti 91 50.1209 10.30516 109 54.9908 8.13372, welch
* 3) Satterthwaite's t-test
ttesti 91 50.1209 10.30516 109 54.9908 8.13372, unequal
HTH.
you could of course follow Bruce Weaver 's SPSS suggestion which appears to me to be slaying a gnat with an elephant gun. If you have the summary stats i suggest your calculator and a basic stat's book will be much faster esp if you type like I do . note that Sat and Welch are pretty close to identical: https://www.google.com/search?q=difference+between+satterthwaitw+and+welch+tests&rlz=1C1CHBF_enUS917US917&oq=difference+between+satterthwaitw+and+welch+tests&aqs=chrome..69i57.56425j0j15&sourceid=chrome&ie=UTF-8) in most real cases.
Best, David Booth
To be fair, David, I did suggest that I (strongly) prefer Stata over SPSS for this kind of problem. Its ttesti command requires only one short simple line of code. ;-)
Bruce Weaver In that sense, you certainly did, but i suspect that he is wedded to SPSS which for the proposed question is IMO a very poor choice. In a similar vein, in R one can proceed exactly as on a calculator but still requires some startup time to know that. IMO calculators are the way to go for this question to emphasize that stat packages are not needed for every little thing. However, that is what he asked for. Best wishes, David
Hi Bruce and David, thanks for offering those suggestions. I'll look at both and see which is the easiest and most effective. Cheers, Nathan
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size