The "results" are uninterpretable.
In statistics, we define the population to which we are going to generalise (e.g. "younger people who have been ill recently"). We then operationalise that into a sampling frame (e.g. "people aged between 15 and 21 years at the time of the survey, who are not in residential care or incarcerated, who had a hospital visit in the last 12 months"). We then cut the sample based on the sampling frame.
Depending on the sample size (which is often affected by funding), we may choose to "over sample" subpopulations in the sampling frame, so we get accurate statistics on them. Some subpopulations can be so small, that you might only get about 10-20 people in your survey, and that's not a large enough number to produce accurate statistics. We adjust for the oversampling by using weights to compensate.
The sampling frame sample is identified ahead of the survey, and the sample is drawn from that (e.g. 500 young people randomly drawn from hospital records that meet the criteria). We then attempt to survey those 500 young people. Based on the number of completed surveys, we calculate the response rate, which is one measure of bias of the results. We then apply our weights to the data so that it is analysed with the sample being as close to the sampling frame as possible (e.g. we might have undersampled boys, based on the overall data, so an adjustment is made for this).
What we end up with are results that are generalised to the entire sampling frame, which are estimates of how everyone in the sampling frame would have answered, had we included everyone. We can do this because we have controlled the sampling frame and the sample.
The Stonewall survey did not do this. It had no sampling frame, and therefore it had no basis upon which to produce accurate statistics. In addition, it is very easy for online surveys like this to be answered over and over by the same person.
So basically the Stonewall report can only be used as an indication of how the respondents to the survey answered. The results cannot be generalised to a population/sampling frame as Stonewall never defined those.
And then there's the questions! Were they leading? Was the survey cognitively tested or field tested ahead of undertaking the full survey?
LOL at the likelihood that happened.
There's a whole bunch of really technical other factors that I won't go into, but that's the gist of how bad that "survey" really is.
And this is the sort of thing that makes me really pissed off at people (present company excluded, obviously) who make out that doing surveys and social science is really easy. Also, strangely enough, the ones pouring the most poo onto this tend to be boys and men.
Tirade over, I promise!