Of the research methods that you mention in this question, the two that can truly be called scientific, and which is commonly used in the social sciences, are surveys and secondary analysis. All of the others are either not very scientific or they are not practical in the social sciences.
In order for a method to truly be scientific, it must be based on objective observation. It cannot be based on subjective interpretations. The problem with most of these methods is that they call for subjective interpretations. This is the case with participant observations and most unobtrusive methods. In many of these cases, social scientists are trying to gather fairly impressionistic “data” about the people whom they are observing. These impressions are not things that can be falsified and they are not, therefore, scientific. The reading of documents will generally fall into this category as well.
Experiments are very scientific. They problem is that they are not available to most of the social sciences. You have tagged this with “sociology” and sociology is not a field in which it is very possible to do experiments. Approval for experimentation on human subjects is very hard to get for sociological purposes.
Surveys and secondary analysis, by contrast, are relatively scientific. Surveys can obtain quantifiable data. Secondary analysis can review the data from many studies and can try to use that data to make conclusions. These are more scientific than observation and they are more feasible for sociology than experiments.