I am collecting data on Quality of life and social capital among African migrants population in Germany. Because of the relatively small size with-in Germany’s migrant population, unavailability of extensive documentation of migrants from this region, the group geographically dispensability and closed networks along ethnic and religious lines – I couldn't come up with a concrete sampling frame and therefore decided to adapted exponential discriminative snowball sampling technique to approximate a random sample.

My first wave of participants were community, network and religious leaders (i.e. African students and professional associations in Germany; Ethnic groups; Nongovernmental organization, Africa missions in Germany) and also members of African groups on facebook (response = 134)

(second wave) I asked the community leaders (who responded  = 134) to share the survey online link/paper and pencil with other Africans who meet the inclusion criteria or provide contact information to potential participant (response = 233)

The third wave will be to ask the 233 respondents for referrers.

Analysis type. 

I intend to do Multinomial logistic regression and path analysis (SES, cognitive and structural social capital and quality of life)

My question....

with this very rough sampling technique (nonprobability) am I still able to perform those analysis and draw a reasonable conclusion? 

What steps can I take to make my sample as representative as possible?

Thanks in anticipation for your replies... 

Best 

Ade