0
$\begingroup$

a conceptual question appear in my mind and i need help, in pratical therms, whats defines the statistical test its my hypothesis and my experimental design ?

Using this example, if i am interrested in analyse the effects of gender and age over scores, i need to collect the data for scores, gender and age all for the same sample from a population, and at this case use of Two-way ANOVA its appropriate, correct ?

Data example 1:

enter image description here

But, what happen if i have the same hypothesis and collect two individual samples, 12 people for age and 12 for gender ? In this case, i need to use One-way ANOVA for the first case and t-test for the second? In this analysis i am creating bias because of a poor design in comparison with the Two-way ANOVA case ?, even if i had a superior N ?

Data example 2:

enter image description here

$\endgroup$
0
$\begingroup$

OK, I think I understand the question.

Yes, if you opt for the second design, you lose the ability to control for gender and age simultaneously. This opens up the possibility of confounding, and no increase in data can rectify that.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks for your answer @Demetri Pananos ! So, all dependes from what i need and what i can measure, in this case when i can measure age and gender, the ideal its measure all togheter and use Two-way. And In the must of the cases when i decide to measure 2 or more dependent variables associated with one independent variable, wich generate 2 or more factors, they need to be bounded and i need to use one test hypothesis that consider all togheter, right ? $\endgroup$ – Leanderson Silva Jul 3 at 1:32
  • $\begingroup$ I will say this: What you measure depends on your research question, and not the other way around. If you think that one variable is a major determinant of an outcome and that the effect of that outcome is not confounded by any other variables, then by all means collect only that outcome. Very rarely, and perhaps only in randomized experiments, is a single factor sufficient to explain variation on an outcome. Though I don't like hard and fast rules, I would strongly recommend controlling for other covariates, especially age and sex. $\endgroup$ – Demetri Pananos Jul 3 at 2:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.