Dear Doaa,

You can use correlation for investigating the relationship and ANOVA for differences. (Post Hoc for sources of difference).

No, It doesn't wrong. But take your type of data into account.

I recommend regression for clarity reasons. You might look at Montgomery Design and Analysis of Experiments available in the Z-library

Best wishes, David Booth

Saludos Doaa. No solo el coeficiente de correlación de Pearson, también ANOVA, Covarianza pero para distintos métodos de tratamiento aplica regresión múltiple, mejor si aplicas el simulador estadístico Risk Simulator, probaras miles de veces el valor de correlación.

Buena suerte

Are you wanting to compare a small number of treatment methods for some variable? If so, how many treatments and what is the variable (and the likely distribution)?

You seem to want to do a correlation? Why? Also, doing multiple statistical procedures to address a single research question can be iffy. It can be p-fishing.

Mohammad Ali Koushesh Vatan

I appreciate your answer. Thank you.

David Eugene Booth

Thank you. I really appreciate your answer.

Jaime Eduardo Gutiérrez Ascón

Gracias, agradezco tu respuesta

Daniel Wright

Yes, I used 3 types of treatment and the dependent variable was (27 parameters for each method). All come under the normal distribution and I want to evaluate their performance statistically.

I really appreciate your answer. Thank you.

Interesting question

Thank

Wow, all 27 are normally distributed. That is really odd. How correlated are the these 27 variables?

Daniel Wright

Yes, the test of normality yielded a result that was greater than (the significance level, 5%). Variables are thus normally distributed.

Is that correct?

Is this the query?

Could you please provide me with a more detailed explanation?

Fatima Boukhlifi

Thank you for your interest.

@Jaime Eduardo Gutiérrez Ascón

Jaime -

What is the "Risk Simulator" please? Where can it be found, and what does it do?

Thanks - Jim

That doesn't mean that they are normal, particularly if your sample is not large. That would be accepting the null hypothesis.

But for the bigger question, since you have three group (a factor or categorical variable) and you are looking for differences among these groups for 27 variables, the next question is what is the structure for these 27 variables. You may want to reduce the number of dimensions. It depends what theories you have for how they relate.

Thank you fot this interesting question