I have different response variables of migrating birds, which I want to investigate via regression:

1.) some kind of speed (distance per hour)

2.) the ratio of restig days versus migrating days

3.) A measure for the "straightnes" of the flight, restricted between 0 and 1 (1 is a perfectly straight fligt route, the closer to zero it is, the more the birds follows a random walk)

Now I would like to choose appropriate probability distributions.

In the fist case, the variance does not inrease with the response variable, hence, I prefer a normal distribution over a Poisson distribution, is this OK?

Regarding 2.) I assume a Binomial-distribution is appropriate, since I can think about this variable as a numer of successes vs a number of failures. Is this allright?

Finally, I have some problems to find an appropriate distribution for 3.). Since this response is restricted between 0 and 1, neither a normal nor a Poisson distribution makes sense at first sight. Any suggestions?

Thanks in advance