Dear Merche

Parametric and NP variables can nit identified by SPSS or SAS or other software.

In 2*2 model you can not use ANOVA because you need more than two mean for analysis.

Check the statistical measure and reanalyzed the data.

Merche, el ANOVA esta correctamente aplicado (como indica Alex). El tema es que para estimar el Cuadrado Medio Residual necesitas al menos unos 6 grados de libertad, es decir te faltan replicas (duplicados o triplicados). Intenta eso y creo que el CMres podria estimarse mejor.

Por otro lado con o sin replica, es posible que NO existan diferencias significativas en tus grupos.

Saludos,

Hola Merche. Estoy en concordancia con lo que te ha contestado Jorge (y porp supuesto Alex). Piensa que el hecho que el resultado de tu ANOVA no te de significativo no significa que deabs optar por un analisis no parametrico. El numero de muestras es determianante para que puedas obtener significacion; aunque por otro lado, no todos los datos de un analisis tienen que dar significativos siempre, de hecho la falta de significacion es un resultado tan valido como cualquier otro explicando tambien tus resultados. De todas formas, y volviendo al pricnicpio de tu pregunta, sabes que antes de aplicar el ANOVA has tenido que seguir unos pasos para comprobar que los datos se ajustan a un analisis parametrico. Si has seguido estos pasos (normalidad, homocedasticidad....) no tiene sentido aplicar un analisis no parametrico.

Espero haberte ayudado

David

Dear Merche.,

I think the number of your participants in each group is important,besides do you check the normality?

Hello everyone,

I've decided not to compute the Two way ANOVA, because groups are really unbalanced. I checked it out using a contingency table.

The database was created by one of my students, and she showed me only the results.

So, what I proposed her is to compute a T-test considering each independent variables separately. She is "really" interested in one variable (control group/clinical group), the other one was sex. She told me that her sample was balanced enough to perform a Two way ANOVA, but that is not true. She wants to analyze if there are differences between these two groups in SCR (quantitative variable). So T-test, may solve the problem.

On the other hand, she showed me several Mixed ANOVA (one factor: group, second factor: repeated measures of EMG) and she had problems with the assumptions. I think the reason of that problem is the sample. What I recommended her was to perform a One way ANOVA for repeated measures only with the clinical group in order to check if there is any change during three EMG measures from the same trial. She has 20 subjects in the clinical group, so, we must be very careful with the assumptions.

I think this is the only solution, because it is no possible to repeat the experiment with more participants.

Thank you very much for your help.

I propose a regression model with two dummy variables plus the interaction betwen them.

Thanks people for helping me. My student decided to start a new research including emotional intelligence among these variables I used to compute the Two-way ANOVA.

Also her database was full of mistakes (with no solution because one on the variables was extrated from a EMG measure which had problems...), so she decided to start again...

I'll keep in mind all your tips you shared! Thank you very much!

In two way of ANOVA you should have two dependent variables and one independent variable. In a 2x4 model is preferable. you have reach you data. and then, apply your objectives because data analysis obtain objectives of study.