To choose a dose-response model in Prism, you need to answer four questions:
Prism offers two sets of equations for dose-response curves. One set assumes X is the logarithm of dose or concentration. The other set assumes X is the dose or concentration. Be sure to pick an equation that matches the data.
Prism offers one set of dose-response equations for stimulation and another set for inhibition. The inhibitory equations are set up to run downhill. The only difference is that the inhibitory equations fit the IC50 ("I" for inhibition) while the stimulation equations fit the EC50 ("E" for effective).
If the curve goes up hill, choose from the set of stimulation equations. If the curve goes down hill, choose from the set of inhibition equations.
Prism offers equations using a standard slope, which have a Hill slope of 1.00 (for stimulation) or -1.00 (for inhibition), and variable slope (fit by Prism). The equations that don't have 'variable slope' in their name assume the standard slope. The standard slope is expected when measuring binding of a ligand to a receptor where there is no heterogeneity or cooperativity. But it turns out that many other log(dose) vs. response curves have the same standard slope.
If your data has more than a few concentrations that lead to a response between say 10 and 90%, then you can ask Prism to fit the slope. If your data provide only one or two concentrations that have a response between 10% and 90%, then your data don't really provide information to define the slope and you'll probably need to choose a model with a fixed slope.
The choice is not straightforward, and there are many situations where it is not clear which approach is better.
If your data have been normalized so the curve runs from Y= 0 to Y=100, you may wish to choose a normalized model. These models don't fit the bottom and top plateaus, but rather force the bottom plateau to equal 0 and the top plateau to equal 100. Only choose a 'normalized response' equation when you have determined the values that define 0 and 100 very precisely. Just because the data have been normalized doesn't mean to have to constrain the curve in that way.
There are many situations where it is not clear whether it makes sense to use a normalized model or not. It is not always a straightforward decision.
Prism has a set of special models used for special dose-response situations: