* These examples are derived from the handouts on 2 sample tests.
* First, we will (erroneously) treat these as 2 independent
* samples. SPSS will give us the results for 2 sample
* tests, cases II and III. 'Equal Variances assumed'
* corresponds to case II (sigma 1 = sigma 2 = sigma),
* 'Equal variances not assumed' corresponds to case III
* (sigma 1 does not equal sigma 2). Note that SPSS uses a
* different formula for Case III d.f. than I have been using.
* I do not know where their formula comes from or if it can be
* considered correct.
DATA LIST FREE / Sex Score.
BEGIN DATA.
1 26
2 30
1 28
2 29
1 28
2 28
1 29
2 27
1 30
2 26
1 31
2 25
1 34
2 24
1 37
2 23
END DATA.
VARIABLE LABELS SEX 'Sex of Respondent' SCORE 'Respondents score'.
VALUE LABELS SEX 1 'Male' 2 'Female'.
T-TEST /GROUPS SEX (1,2) /VARIABLES SCORE.
* Now we will treat this as a matched pairs problem. SPSS will
* give us the results for 2 sample tests, case 4.
DATA LIST FREE / HScore Wscore.
BEGIN DATA.
26 30
28 29
28 28
29 27
30 26
31 25
34 24
37 23
END DATA.
VARIABLE LABELS HScore 'Husbands score' WScore 'Wifes score'.
T-TEST /PAIRS HScore WScore.
* Finally, using different data, we'll show what happens when Model II is used
* with a large data set when Model V is technically called for. Note that results
* are virtually identical to what we got when we hand-calculated the results
* for case V. Later, we'll show another and more exact way to get SPSS to
* handle case V.
DATA LIST FREE / Group Recover Freq.
BEGIN DATA.
1 1 75
1 0 25
2 1 65
2 0 35
END DATA.
Value labels Group 1 "Group A" 2 "Group B"/
Recover 1 "Recovered" 0 "Didn't Recover"/.
Weight by Freq,
T-TEST
GROUPS=group(1 2)
/MISSING=ANALYSIS
/VARIABLES=recover
/CRITERIA=CIN(.95) .