| Question |
Comment |
| Is the logEC50 reasonable? |
The EC50 should be near the middle of the curve, with at least several data points on either side of it. |
| Are the standard errors too large? Are the confidence intervals too wide. |
The SE of the logEC50 should be less than 0.5 log unit (ideally a lot less). |
| Is the value of BOTTOM reasonable? |
BOTTOM should be near the response you observed with zero drug. If the best-fit value of BOTTOM is negative, consider fixing it to a constant value equal to baseline response. If you know where the bottom of the curve should be, then set BOTTOM to that constant value. |
| Is the value of TOP reasonable? |
TOP should be near the response you observed with maximal concentration drug. If the best-fit value of TOP is not reasonable, consider fixing it to a constant value. If you know where the top of the curve should be, then set TOP that constant value. |
| If you used a variable slope model, are there enough data to define the slope? |
If you asked Prism to find a best-fit value for slope, make sure there at least a few data points between 10 and 90% . If not, your data don't accurately define the slope factor. Consider fixing the slope to its standard value of 1.0 |
| If you used a model with a standard slope, does the data appear to be steeper or shallower? |
If the data appear to form a curve much steeper or shallower than the standard dose-response curve, consider fitting to a model with a variable slope. |
| Does the curve appear to be biphasic? |
The standard dose-response models assume that the curve is monotonic. If the curve goes up, and then down, you'll need a more complicated model (beyond the scope of this manual). |
