Student t-test vith ViSta - How to test for equal/unequal variance between groups ?

[Michael]

Hello,

An assumption of t-test for independent samples is that groups or conditions have equal variance. When I use the UNIVARIATE procedure in ViSta, I only get the t-test that assumes that variances are equal in the two groups.
- Is there a way to get a formal test, foe example something like the t' test (Welch t test) that modify the degrees of freedom in order to get a correct inference in cases where variances are unequal between groups ?
- Is there a way to test for equal / unequal variances in the groups (e.g. SPSS shows Levene's test for equal variances) ?

If not, how do you procede to verify the assumptions of the Student t-test ? Do you use only visual plots (e.g. like box and whiskers ?)

Regards,

Michael

[Pedro Valero]

Sorry for the late reply. I do not know why but I did not see this message until today.

Yes, there is not a test for unequal variances programmed in ViSta. I might add it if the formulae are not entangled if you wish. Do you have a good summary I could use (I know I am being lazy but I am quite busy these weeks...)

About, diagnostics for unequality of variance, I personally rely more on visual inspection of the data. This conforms with statistics books (Moore or De Veaux et. al) which warn of the lack of robustness of the homogeneity of variance formal tests like Levene. These books recommend graphical tests such as boxplots.

However, I see in the output of ViSta that there is a test of this (see last line), Is not this what you wish?

T-TEST FOR TWO INDEPENDENT SAMPLES:

Null hypothesis:        Mu1 - Mu2 EQ     0.000

Alternative hypothesis: Mu1 - Mu2 NE     0.000 (two tailed)

SAMPLE STATISTICS:

H............ N =       3 Mean =    0.667 StDev =    2.082 Var =    4.333

C............ N =      73 Mean =    7.082 StDev =    4.287 Var =   18.382

Sample Differences... DiffMean =   -6.416 StErr =    2.499 Var =   18.002 (Pooled)

SIGNIFICANCE TEST

Test Result.. T =  -2.567 For df =  74.0 p = 0.0123

TEST OF HOMOGENEITY OF VARIANCE: 

Test Result:  F =   4.242   df = (72 2) p =   0.1288