I can use random sampling (probability) for selecting org. and for selecting respondents I can go for convenience (non-probability) sampling

Nuray -

Convenience sampling isn't really good for inference to the population. For inference with accuracy you might estimate, you need a probability-of-selection-based (design-based) approach, or a model-based approach, or they can be combined, but this is not what I'm referring to. Still, it might be feasible if you had regressor data on everyone in your convenience sample, and all in the selected clusters, but that would be something new, mixing up definitions of variance, I think, and convenience sampling isn't necessarily a good way to handle the modeled part.

Are all of the respondents within an organization doing the same job? If you considered this a different sample for each position, then for any unique position, say CFO, you have a random sample of CFO's if you request the CFO from each organization selected to reply. If this is a voluntary survey, and you have quantitative data from CFOs, you can use imputation, if just a few are nonrespondents. But you have to be careful of introducing bias or artificially reducing variance, depending upon what you do. Best to document any assumptions.

For positions for which there are multiple people, your problem is tougher. For quantitative results, it gives you unknown bias, and variance cannot be estimated either. For qualitative results, I guess you have to document what you did, and again hope your results aren't extremely biased.

Best wishes - Jim

Dear Nuray,

Effectively, you can use more than one procedure in the same study. This generally call multistage sampling.

You can do, but you should be specific about why you are choosing each technique of sampling for each phase or each level. Through your description, I believe you should differentiate between the unit of analysis and unit of observation.

Unit of analysis refers to the entity about which the data are collected (e.g., firms, groups, industries) to answer the research questions. In other words, the units of analysis are the entities that the researcher will be able to communicate their results at the end of the study. Unit of observation, on the other hand, refers to the individual source from which the data are collected from; and especially common in quantitative data analysis (e.g., econometrics). In other words, units of observation are the entity that provides data to be analyzed to study the unit of analysis.