hi,
i want to compare the perceptions of teachers and learners regarding critical reading with a questionnaire which consists of 8 factors, what statistical test should be used to find the difference across theirs? t test /anova?
should such a statistical test be conducted with each 8 factor?
it should be noted that the questionnaire has 8 constructs and 38 items . if i want to compare two groups' perception with respect to each construct what statistical test should be used. should i carry out 8 t tests for each construct or anova + posthoc test. by the way if i use 8 t test s what post hoct test should be used/?
A t-test is a special case of an ANOVA; i.e., for a one-way design with only 2 levels. Most statistical software packages for t-tests yield more detailed/complete results for t-tests than for a corresponding one-way ANOVA (e.g. tests for homogeneous vs. heterogeneous variances). If you also want to compare the 8 factors, you could run a two-way anova with repeated measures or a MANOVA.
Dear Rokhsareh
You must test your data for normality distribution and the homogeneity of variance, in case of providing these assumptions, then you can use T-test for each item of the questionnaire.
Good Luck
thanks jos and khalid,
it should be noted that the questionnaire has 8 constructs and 38 items .
if i want to compare two groups' perception with respect of each construct what statistical test should be used. should i carry out 8 t tests for each construct or anova + posthoc test.
by the way if i use 8 t test s what post hoct test should used/?
You can use t-test assuming your perception scores are continuous.(ANOVA will also give the same results, BTW). It would be more convenient using Mann Whitney U test, if scores are obtained from Likert items by some combinations. Since there are only two groups no post-hoc tests are available for a t-test or Mann Whitney U.
As you have two groups 1.Teachers 2. learners so you need to apply t-test.
If your 'constructs' are based on the means (or the sums, if there are no missing values) of a number of the 38 items, then you can use these as interval scales. Yes, do test for normality. Yet, a t-test is known to be quite 'robust' against violations of assumptions of normality and homogeneity of variances. If you are interested in the differences between the 8 'constructs' (repeated measures, i.e. a within factor) and/or the interaction with these and the perceptions conditions (between group factor), then you could use a two-way repeated measures ANOVA (for mixed, i.e. between x within designs) . If you prefer nonparametric tests, there are such tests for mixed designs, for instance the nparLD R package or nparLD SAS macros.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA