How is plot in scaled estimate (DOE analysis) plotted, what type of plot is it?

[#1]

Hi,

I want to understand what type of plot is used in JMP DOE report : Scaled estimate table and sorted estimate table. can someone help me to understand it so I can use it correctly?

Thanks

[#2]

Go to Help in the JMP menu.  Search on scaled estimates in the JMP Documentation Library.  Page 150...

[#3]

I checked, but it didn't explain the plot in the scaled estimate. There is not header for the plot column. It don't looks like bayes or pareto plot. Can someone helps? When right click on the scaled estimate table>columns, it only shows "plot estimate". what type of bar chart is that and why it has 5 vertical lines (including zero) on both negative and positive side?

 

[#4]

Hi @dadawasozo,

That's a bar chart to help you quickly visualize the values of the estimates relative to each other. The largest bar is the largest value in the scaled estimates column and the other bars are scaled accordingly. If there are negative values the bars are centered at 0 and the negative values move to the left.

It doesn't have any complex interpretation, just use it to quickly find large values and small values.

[#5]

I have a follow-up question. I search through document and found that the scaled estimate are scaled to have a mean of zero and a range of two. Does it mean take the z-score and divided by 2? Why divided by 2?

[#6 · Accepted solution]

Using scaled estimates in the DOE platform, the scale is to normalize the actual values of the factors impacting the size of the coefficients.  For continuous variables the estimate is centered on the mean and the takes the effect divided by 2.  Think of it this way...take a hypothetical model:

Y=b0 + b1X1 + b2X2 + ...+ e

where b0 is the Y intercept (the mean of the data set) and b1 is the slope for X1.  The slope (rise/run) is the effect of the factor divided by the change in x (for a 2 level design, this is 2).  When you put an actual value in for X1, you don't get the entire effect of X1, only the change from the mean (which is half the effect).

[#7]

thanks for explanation.

I found that my parameter estimate and scaled estimate have the same value for all effects. I thought these two estimate shouldn't have the same value. Under what situation that these two estimate will be same?

[#8 · Accepted solution]

When you use the coded levels (by default in JMP) for the factor levels (e.g., -1, 1 for 2-level factors)