[rahimnathwani]

The article is not based on a random sample of homeless people. If you read the article, and ideally the article linked within, you'll notice that both the control and treatment groups are from a set of people who have been able to maintain ongoing relationships with 'buddies', in person or on the phone.

The linked study claims to be based on a representative sample of homeless people, but the sampling approaches detailed on page 14 don't give me confidence that this is true. This part is particularly problematic: "This process continued with participants referring us to members of their communities who then referred us to others."

The final weighted sample might be representative based on some demographic criteria (e.g. ethnicity, age) but weighting it to achieve that doesn't magically make it representative of the overall homeless population in other respects.