Excuse me because the attach paper is in spanish, but you need to make a factorial design and then do with minitab (statistical software) a fractional design: Finally, with the experiments see the influence of each variable over the response. If the response is horizontal, this variable not influence over the response.

Dear Waleed,

Simply apply PCA at your all independent variables. Select those variables wit higher cosine value. It will make easy to simplify the model.

The variables with higher cosines can be taken as control, because these will be harvesting maximum variability.

Regards,

(JAMIL)

Could you explain tour method? is it scientific method? do you have any published paper about that PCA?

Hi Waleed,

I would advise you to refer carefully to the literature. It's quite similar to the hypotheses development process, where we use arguments found in the literature to predict relationships and formulate hypotheses.  In the same manner, the literature can also provide strong justifications for using some variables as IVs, some as control variables, while others may be moderators, mediators and DVs. 

Hope this helps!

Dr Aida Idris, University of Malaya, Malaysia

how much is a lot ?

18

I agree with Muhammad.

PCA: Principal Component Analysis and Factor Analysis are the methods that you may use. Moreover, these methods can be very easily applied using the open source r.

Regards

Nethal

Waleed Fraij,

You may use t-test analysis to remove irrelevant features and then use some feature selection technique to select best set of independent variables. Please follow these paper:

https://www.researchgate.net/publication/303011477_Quality_Assessment_of_Web_Services_Using_Multivariate_Adaptive_Regression_Splines

https://www.researchgate.net/publication/299328355_Hybrid_functional_link_artificial_neural_network_approach_for_predicting_maintainability_of_object-oriented_software

https://www.researchgate.net/publication/292677172_Maintainability_prediction_of_web_service_using_support_vector_machine_with_various_kernel_methods

Conference Paper Quality Assessment of Web Services Using Multivariate Adapti...

Article Hybrid functional link artificial neural network approach fo...

Article Maintainability prediction of web service using support vect...

Dear Waleed, a good way is PCA but you need continuos variables. You can also perform factorial analysis with different kind of variables (categorical, binary), but it is little different from PCA. I suggest you to avoid factorial analysis if you are not expert and focus on regression. Of course, it depends from how many variables you have. Use backward selection to avoid useless variables. Use interactions to consider individual characteristics as moderator variables. Use simple slope analysis to understand the interactions.