Dear Zol,

I think the issue is more to do with which test is appropriate rather than which software to use. The Fisher's exact test works well with small samples. It is available in SPSS under the chi-squared test when you have a 2 by 2 table.

You can use any statistical software which can handle non-parametric tests or exact tests. 

The major question is if the assumptions for the test you want to use hold true.

Is supplier performance a continuous variable? Do you have 4 repeated measures of it? Your independent variables are categorical with 3-4 categories. Is that correct?

R, SPSS, SAS, Minitab, Statistica, and a long list of others should work. All of the programs will give poor output if not used properly. With the number of variables that you have described (4 dependent and 12 to 16 independent) it may be that you have insufficient samples to develop any model. That said, if you can simplify to "supplier Performance" as one variable and four independent variable, that could work. If most of the variables are categorical, you may just be out of luck. It could get even worse if there are missing data.

Thank you guys all  for answers and advises!! Really appreciate it!  I have 3 independent variable (abbreviation: SS, SI, SD), where: SS have 5 measurements (total of 14 variables); SI have 3 measurement (total of 6 variables); SD have 4 measurements (total of 9 variables). Basically, total of 29 questionnaire items for independent variables. Dependent variable (SP) consist of 4 measurement criteria with total of 12 variables ( questionnaire items). All variables are ordinal with 5 point Likert scale (from strongly agree to strongly disagree). Sample size is 30 companies. i have another dependent variables, which is MP, but I willing to take that out cuz if i include it SP will be like a mediator. Unless, there are method  I can check relationship between SP and MP separately....

@Peter i think you right, I need to choose, which analysis is appropriate to use.... So basically, I think I ll use SPSS.... After reading many things on analyzating it seems like to find the impact of independent variable on dependent variable for my research, the correlation and regression analysis seem appropriate.... but here comes the question???

1. In correlation analysis should i take Pearson, Spearman, Kendall coef??   

2. For regression analysis it seems like Ordinal logistic regression is right.... Or I can use linear regression as well?

3. And to do the regression analysis what should i take to combine independent variables and dependent variables? Sum or Mean of the variables??

4. I think with low sample size I will try to use Bootstrap on everything.....

5. I am trying on Amos as well but not sure if it can give any valid result ( will try to do bootstrap also but not sure which one I should choose there)...

P.s. My reability test is not bad (everything is above 0.8) but when I am doing validity test KMO is really low like 0.5  

thank you, 

Zol

Help.

You surveyed 30 companies using a questionnaire of 29 questions.

You then use 14 of these questions to calculate a new variable SS.

What are the 5 measurements? Does this mean that only 5 of the 30 companies provided enough information to enable you to calculate SS?

No... what I meant was that there are 3 independent variables (Supplier Selection, Supplier Involvement and Supplier Development), where SS have 14 questionnaire items which measures importance of Cost, Buyer Need, Capability, Responsivenes and Quality&delivery service (5point likert scale questions), SI have 6 questions, which measures Supplier involvement in new product, quality and trust ((5point likert scale questions), SD have 9 questions, which measure companies' effort, responsiveness, long term relationship, management involvement (5point likert scale questions). There is a Supplier Performance, which is dependent variable, which include 12 questions to measure companies' supplier performance(quality, cost, service, flexibility). .. So in general, my hypothesis is like: SS impact/ have relationship  on SP;  SI impact on SP; SD impact on SP... I have 30 companies, which answered my questionnaire... they are all pretty big companies and I asked mostly, CEO, general managers or head of supply departments (no missing value)...  

Ok, so in your previous post you said that:

SS had 5 measurements

SI had 3 measurements

SD had 4 measurement criteria

SP had how many measurements?

MP is a separate variable measured directly (not a composite of several questions).

What are these measurements? Or is this something that we should ignore and move on?

You are saying that they are not replicates. They are not variables. It is very easy to give bad advice if one does not understand the problem.

Run a correlation analysis for all 5 variables (SS, SI, SD, SP, MP). If the correlation between all of the independent variables > 0.9 then state that fact and only work with one variable. If the correlation between the independent variables < 0.2 try building a multivariate model. SP=SS+SI+SD. Also try SP=MP, and based on the outcome consider MP=SS+SI+SD. Check for significant interaction terms (SS*SI, ... ,SS*SI*SD). While your measures for a specific variable are discrete, it is possible that the combined measures (SS, SI, SD, SP) can be considered continuous. Check the residuals for normality with something like a Q-Q plot. Also plot residuals against the predicted values. In the plot of residuals you are looking for patterns: stripes, waves, increasing dispersion about zero with increasing predicted (or the reverse pattern). Clumps (80% of the residuals are bunched on the left side of the graph, there is a gap, and then the remaining residuals are bunched at the right side of the graph. Such patterns suggest that the model is not fitting well. Transformations can help, as can selecting a different model.