Nonlinear regression when you entered error values directly
Prism lets you enter XY data either with side-by-side subcolumns for direct entry of replicate values, or with labeled subcolumns for entry of error values computed elsewhere. What happens when you fit such data with nonlinear regression depends on exactly what you entered.
Case 1: You'll get the same results as if you had entered raw data
When you enter data in the formats listed below Prism takes all that information into account, and you'll get results identical to what you would have gotten had you entered the replicate values directly:
- Mean, SD, and n
- Mean, SEM and n
- Mean, %CV and n
Notes:
- It is essential that you enter the sample size (n) values as well as SD, SEM or %CV. If you omit n, Prism will ignore the SD or SEM values when fitting a curve.
- It matters that you formatted the table so the subcolumns are labeled to match what you enter. The SD and SEM are different, so make sure you set up the table correctly. If you enter SD values in a subcolumn labeled SEM, the results will be wrong.
- On the Weights (Prism 5-7) or Methods tab (Prism 8) of nonlinear regression, you can choose to have Prism fit the mean values only and ignore the sample size and error values you entered. If you make this choice, the results will not be the same as they would have been had you entered the replicate values directly.
Case 2: Prism fits only the mean (or median) values you entered, but ignores error values
When the XY table is formatted in the styles listed below, Prism cannot account for the error values you entered, so fits only the mean or median values and ignores the other subcolumns:
- Mean and SD (but no n)
- Mean and SEM (but no n)
- Mean and %CV (but no n)
- Mean (or median) and upper/lower error limits
- Mean (or median) and plus/minus error values
These formats simply don't provide enough information. All Prism can do is fit the mean (or median) values you entered.