Data Analysis method based on individual respondents or firm

[amir01]

Dear All
Hi
I have gathered 257 questionnaires from 104 firms. What I had in my mind was that I can not transmit all of my 257 data-set to Smart PLS 2. as my statistical population is in firm level not individuals. Because my theoretical model is to measure innovativeness at firm level. So, it means that I received either 1 response (Questionnaire) from one firm or 4 , 5 ,10 from other companies. Each firm responded differently in terms of the number of respondents. I would like to ask if I must mean and getting average of respondents of one firm as various persons and individuals replied to my questionnaire? Or if there is another way of doing so. What I have done up to now is just getting the average or meaning of those firms with different numbers of respondents. Is that way can be considered correct or there's another way of calculation. In the meanwhile, based on the level of firm respondents and getting average, the number of my data set has decreased so that it had somehow some bad impacts relations of my IVs and DVs as well as moderating variables.

Please kindly help me and assist me in this regard how to deal with this issue?

Your kind help is appreciated in advance.

Best Regards
Amir

[jmbecker]

Generally, you dataset is not really optimal. One would usually do a multi-level model for such data, but (1) this is not possible in PLS as this requires Maximum Likelihood estimation and that is not what we do in PLS, and (2) you need at least a couple of observations for each firm (i.e., at least 3-5 observations for every firm).
As you want to analyze the model on the firm level, it would also not be correct to include all individuals (as this would analyze the model on the individual level). Thus, the averaging approach is probably the only one that you could use, although it is less than optimal.