Questions about Pareto Plot of Estimates

[#1]

Dear community,

I have three short questions about Pareto Plot of Estimates.

1. How is t Ratio calculated?

2. How can I interpret t Ratio?

3. The solid black line is cumulative of t Ratio; however, how can I interpret this line?

q_pareto.PNG

Thank you!

[#2]

1. t ratio is the parameter estimate/ standard error of the parameter estimate.

2. t ratio is but one 'statistic' that can be used to evaluate the significance level of a given parameter estimate.

3. The curve is just the sum of the absolute values of the t ratio, it's main purpose is to give you an idea of the relative explanatory power of each estimate wrt to the others.

[#3]

Thank you for your reply.

Is it possible to change from t-ratio to estimated effect instead?

Sincerely,

TK

[#4]

I would generally advise against trying to intentionally change names for statistics, tests, and general nomenclature in JMP to other 'words'. In general JMP developers go to great lengths to use generally accepted or recognized words/terms in analysis platform dialogue windows, reports, and general user interface applications because these words have recognizable and often standard meaning to both producers and consumers of the information contained in the reports or JMP user interface items. Plus all the JMP documentation ties to these words...not words that are replaced by an individual user.

[#5]

You would not want to use the estimated effect for the pareto because the effect is influenced by the scale of the x variables. For example, suppose temperature on a scale from 300 to 500 and pressure on a change from 1 to 2 are both in the model. If they have equal impact, the parameter for temperature must be much smaller because of the larger scale. The t ratio removes that effect.

[#6]

I understand your points. However, are there any ways that I can see let say which parameter has the greatest effect?

Sincerely,

TK

[#7]

The Pareto Plot is one of the ways to assess statistical significance and decide which effects are not null. Examine the Parameter Estimates for importance. If the experiment was designed by JMP, then JMP saved a column property for each of the continuous factors called Coding. The estimates are based on the coded levels of the continuous and categorical factors. The estimates can be used directly to assess the magnitude of their associated effect.

[#8]

This script might be useful. It creates an Effects Plot.

[#9]

Thank you very much!

[#10]

Thank you for a very good effects plot.

Would it be possible to know which equation you use to calculate effects?

Sincerely,

TK