Hi Ria Umair,

Following link may be helpful for your question in determining sample size.

http://www.surveysystem.com/sample-size-formula.htm

Good Luck!

Terans

http://www.surveysystem.com/sample-size-formula.htm

You need to identify the factor with the greatest variability. Quantify the variability in that factor, and use that for your sample size calculation.

Often sample size calculators are useful to get people thinking more carefully about their sample size. However, their accuracy is dependent on factors that we do not know in advance. Sample size calculators often do not include factors like non-response rates, or ethical considerations (mostly an issue for experiments on vertebrates, especially people).

There are online sample size calculators: GPower is one. http://gpower.hhu.de/

You could play with this calculator by entering several different guesses for variability and seeing how sample size changes.

You could look at the scientific literature and find 10 to 20 articles that have described research similar to yours. Document the sample size they used and their justification. Follow their lead.

You should consider what kinds of analyses you will run. Are univariate tests the greatest thing on earth, or will you also use multivariate methods? Maybe both approaches?

The simplest approach is to do as many replicates as you possibly can considering your budget, time, and all applicable ethical issues. If you are very good at collecting data, you might be faced with having to deal with the problem that you show a significant result that is so small as to be meaningless. This is a far better problem than having too little data and being forced to develop arguments that your results are not an artifact of random noise caused by sampling. Also, small sample sizes will greatly restrict your options in data analysis because you will not have sufficient degrees of freedom to build the model.

Another simple approach would be consult a statistician from your institution who is familiar with these sorts of experimental designs.

Dear Ria Umair,

There are so many feature ranking techniques available in literature such as Chi Squared test, Gain Ratio Feature Evaluation, OneR Feature Evaluation, Info gain feature evaluation etc. You may consider any one of them for ranking the features.  I suggest you to use WEKA tool in place of SPSS.

If you need some more details about these techniques please mail me ([email protected]).

Please refer this papers:

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

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 Maintainability prediction of web service using support vect...