Adjust formula to maximum or minimum

[#1]

After DOE result analysis the regression model frequently predicts values that are higher or lower than an expected range of the output Y e.g. for yield max = 1 and min = 0. Is it possible to adjust the formula so that Y can be kept between specified limits?

[#2]

The linear regression model assumes that the errors are independently and identically normally distributed. It also assumes that the range of the function is negative infinity to positive infinity. You could clamp the model prediction.

temp = model( X );
temp = Maximum( 0, temp );
Y = Minimum( temp, 1 );

Model(X) is the saved regression model.

Alternatively, you could model Logit( Y ); This treats Y as a response with a range of [0,1]. Your model will be saved with a series of column formulas. The one for probability is your response.

[#3]

Hi Marc, can you apply your first option = temp to attached data table; the model prediction formula is Selectivity (%). Thanks

[#4 · Accepted solution]

It was intended as a suggestion, not an exact formula, so I edited it. I added it as a new data column. All of the predicted values were within the [0,100] interval, so nothing changed.

Note that you asked about [0,1], but the data are actually [0,100]. Modify the formula as necessary.

[#5]

Works fine, thanks a lot Marc!