150 likes | 340 Views
Different Kinds of T-tests. One sample t-testWhen population variance is not knownTwo sample t-test for independent samplesCompare two sample meansSamples are independent of each otherTwo sample t-test for dependent samplesCompare two sample means Samples are paired / dependent. One-Sample T-
E N D
1. Research Methods I T-tests
2. Different Kinds of T-tests One sample t-test
When population variance is not known
Two sample t-test for independent samples
Compare two sample means
Samples are independent of each other
Two sample t-test for dependent samples
Compare two sample means
Samples are paired / dependent
3. One-Sample T-test One-sample t-test when µ is known but s is not known.
For the one-sample t-test sM is being substituted with sM (standard error with estimated standard error)
4. T-distribution The sampling distributions of z and t are different.
For z-tests the sampling distribution of sample means is a normal curve
For t-tests the sampling distribution changes according to sample size (above n=120 almost like z distribution)
5. Degrees of Freedom Degrees of freedom are the number of scores in a sample that are free to vary. Because the sample mean places a restriction on the value of one score in the sample, there are n-1 degrees of freedom for the sample
The critical value for t depends on the degrees of freedom (df)
Degrees of freedom depend on the sample size
df = n-1
6. Two-sample t-tests In two-sample t-tests two sample means are being compared.
What is the probability that the difference in the means that is observed is due to chance?
Population parameters are no longer known
Normal distribution of population must be assumed unless sample size is very large
There are basically two kinds of t-tests
Dependent
Independent
7. Dependent vs Independent Samples Dependent samples
Repeated-measures, or within-subjects design
Members of one sample are matched or paired in some way with members of the other sample
Independent samples
Independent-measures, or between-subjects design
Members of one sample are in no way related to matched to members in the other sample
8. Independent Samples t-test Assumes that composition of one sample is independent from composition of other sample.
Both samples reflect different populations
H0: µ1= µ2
H1: µ1? µ2 or µ1< µ2 or µ1> µ2
Sample size (n), sample mean, and sample variance have to be known for each sample
If two random samples are collected and one has had a treatment exposure only the treatment exposure should make the difference between the two groups.
9. T-test for Independent Samples T-test for independent samples is based on distribution of differences between means.
It is like the t-distribution except using the differences between means
In order to calculate t we get the difference between the two sample means, and divide it by the standard deviation of the sampling distribution (the standard error)
However, standard error for this distribution is based on differences of means and is therefore different.
10. T-test for Independent Samples 1. Write out null and alternative hypothesis for the original problem.
2. For each sample determine its sample size, its mean, and its variance
3. To determine which t formula to use for the F test for homogeneity of variance
4. Perform the appropriate F test
11. T-tests for Independent Samples The standard error for our distribution is based on variances of two samples.
If both variances are close in magnitude we can calculate the pooled estimate of common variance. This is based on a weighted average of our two sample variances.
If both variances are not close in magnitude we have to use a different formula for the t-test
12. F-test for homogeneity of variance The F-test for homogeneity of variance tells us whether the variance between both samples is similar or different
Write out the null and alternative hypothesis for the F test
Calculate F and its two degrees of freedom
Compare the F value with Fcritical
If your F value is smaller than Fcritical then assume equal population variances.
If your F value is larger than Fcritical then assume unequal population variances.
13. F-test for homogeneity of variance H0 : variance of sample 1 is equal to variance of sample 2
H1 : variance of sample 1 is not equal to variance of sample 2
F tests are always non-directional
F = larger of the two sample variances divided by the smaller of the two sample variances
Degrees of freedom for both numerator and denominator: df = n – 1
F-table gives critical values
n1 = numerator degrees of freedom
14. T-test for SES 15:04 Wednesday, October 22, 2003 4
The TTEST Procedure
Statistics
Lower CL Upper CL Lower CL Upper CL
Variable group N Mean Mean Mean Std Dev Std Dev Std Dev Std Err
ses 53 3.7184 4.1698 4.6212 1.3746 1.6377 2.0263 0.225
0
ses 46 3.7534 4.2391 4.7248 1.3565 1.6355 2.06 0.2411
1
ses Diff (1-2) -0.724 -0.069 0.5853 1.4352 1.6367 1.9045 0.3298
T-Tests
Variable Method Variances DF t Value Pr > |t|
ses Pooled Equal 97 -0.21 0.8340
ses Satterthwaite Unequal 95.1 -0.21 0.8340
Equality of Variances
Variable Method Num DF Den DF F Value Pr > F
ses Folded F 52 45 1.00 0.9981
15. SAS for Paired T-tests data new;
set work.julia20;
diff = final_exam - mid_term_exam;
run;
proc print;
var diff;
run;
proc means n mean stderr t prt;
var diff;
run;
16. The MEANS Procedure
Analysis Variable : diff
N Mean Std Error t Value Pr > |t|
-4.0000000 1.1742180 -3.41 0.0059
For dependent sample t-test we run a proc means with the options n, mean, stderr, t, and prt
These statistics will be computed for the difference variable
T will give the t-value and its probability, testing the null hypothesis that the variable DISS comes from a population whose mean is zero.
The mean gives the average difference score. If p<.05 we can say that the two groups are significantly different from one another.