Responses to your queries:

a. Take average of the indicators to compute a composite variable.

b. Try SmartPLS, it is effective tool for small data set.

Cheers!

Multiple correspondence analysis is typically for analyzing nominal data - your individual item data is ordinal and your "average" item score is probably going to be a decimal (unless you round to the neared Likert score option). How would you go from these scores to nominal indicator 0/1 categories for the indicator matrix?. You're losing a lot of information content along the way.

Checking I've got your data right: you have 40 items with each one rated on several indicators using an 11-point Likert scales (-5 - +5), with a lot of missing data. Total of 20 respondents.

Not sure what your analysis aims to do, but perhaps check into multicriteria outranking methods. Quoting from the Promethee-Gaia website: "Outranking methods provide decision makers with an alternative to the aggregation methods that are popular in the US and the UK. Typically they require less information from the decision maker(s) and allow to keep closer to the actual decision problem."

There is an excellent website here: http://www.promethee-gaia.net and their Visual Promethee analysis software is available free to students.

Alternatively, perhaps a cluster analysis method would be more appropriate to group technologies according to how similarly they performed on the various indicators.

Wondering too, does it matter if a technology gets an average score of say 0 on an indicator because everyone rated it 0, or if ten people rated it -5 and ten rated it. Would that "mean" the same to you in terms of interpreting its overall score?

Likewise, would a composite score of say 5 be equivalent if it came from 5 indicators rate(d 0, 0, 0, 0, 5) versus being rated (1, 1, 1, 1, 1)? A lot of useful nuance can be lost by creating summary scores and then summing unlike things into a composite score - it can end up implying that the criteria are all essentially indistinguishable and are measuring the same underlying construct.

IF this is partly or wholly so, perhaps check out a bit of reading on psychometric scale construction, or inter-rater reliability - depends if you expect things to be rated very differently by your 20 participants or if there is an assumption they'd all value things in a similar way.

Dear Warren , 

Thanks for your elaborate response. Actually this scale is transformed from a quantitative understanding. For example 

I have a survey data where ratings have been given between -5 to 5.

5 means: a practice increases water level by 100%

4 means: a practice increases water level by 80%

3 means: ,,                    ,,                    ,,       by 60%

2 means: ,,                   ,,                    ,,       by 40%

1 means: ,,                   ,,                    ,,       by 20%

0 means: no change

-1 means:   decrease by 20% 

and so on till -5 and its corresponding decline of -100%

So for every score I have a defined quantitative understanding. Hence, one of the researchers on "researchgate" pointed out that my data is quantitative and -5 to 5 is rather a transformation. Hence, its not ordinal. 

Now my plan is not to compute a composite score, but a more disintegrated analysis. I am planning to conduct a cluster analysis rather to characterize each technology with some performing indicators like cost, yield, feasibility etc. Can I apply PCA on these mean responses with K-means clustering ? 

@Warren , another query. In case I do not compute a mean response, can I proceed with multiple correspondence analysis (to do clustering). I have 43 technologies which are to be evaluated on 10 performance indicators. 

Ag-tech is not my subject area so I'm hesitant to comment much further as there may be domain-specific considerations that would inform the most appropriate answer.

I would agree with the comment that your "ordinal" scale is actually quantitative in an important way, but since it is about relative increase/decrease rather than absolute change, your transformation may still make the most sense for comparing items - would depend on the specifics of your study. IMHO interpretability of the result in light of the research question should be a key consideration for formulating the analytic approach.