Normal quantile plot for different variables

[#1]

Hello, 

I have data which correspond to different cases (resistance values for different structures). I want to make the normal quantile plot for each case and see all cases in the same graph (each case with different marker or color, for example) in order to observe directly the differences and make comparisons.

Is it possible? I try grouping by structure in the option "by" but I obtain one different graph for each case.

Kind regards,

G.

[#2 · Accepted solution]

This result is possible in the Oneway platform. I assume that you have the data in one column and the group indicator in another column. Select Analyze > Fit Y by X. Select the data column and click Y. Select the column with the group indicator and click X. Click OK to launch Oneway.

Click the red triangle next to Oneway and select one of the commands to plot the normal quantiles.

Note that you can also select Rows > Color or Mark by Column and select the grouping column. This command (by default) will assign the same color to markers as used for the lines for each group.

[#3]

Hi - related to this solution: How can we create plots like this for multiple levels of nested groups? Let's say I have a grouping A (e.g. wafer ID) and a grouping B (e.g. test structure), that will nest within group A.

[#4]

I must apologize, and please do not be offended, but why would you want normal plots for a nested sampling plan? One of the most important aspect of such a plan is the rational series in which the data was acquired (e.g., time). When you put it on a normal plot, you lose the series. Now I think normal plots for DOE analysis are fantastic, but the time series in DOE is secondary to the factor effects.
“Analysis of variance, t-test, confidence intervals, and other statistical techniques taught in the books, however interesting, are inappropriate because they provide no basis for prediction and because they bury the information contained in the order of production. Most if not all computer packages for analysis of data, as they are called, provide flagrant examples of inefficiency.” Deming, W. Edwards (1975), On Probability As a Basis For Action. The American Statistician, 29(4), 1975, p. 146-152
Use variability chart, control charts (Range to assess the test structure consistency and Xbar to compare the test structure variation to the Wafer ID. Is the test structure within wafer consistent and if so which source is greater: the test structure or the wafer-to-wafer variation). and nested components of variation for quantitative (which you can get using the variability chart (options Variance Components).

[#5]

Hi - thanks for the response. I'm not sure why you are discounting the use of a normal quantile plot for nested groups. I would like to put all the groups together for visual comparison on the same plot. Plots like this are common to compare the distribution of a certain parameter across different experiments.

Here's a visual of what I'm trying to achieve:

No offence taken and none intended.

[#6]

It's not that I am discounting it, I'm just suggesting you are throwing away information. My quote from Deming is the point. When you simply look at the distribution and outliers you miss information about the time series. How do you compare the 4 combinations? How do you separate the nested effects? How different must they be to be significant? Shewhart invented control charts for just this purpose.

[#7]

Not sure why you are assuming there is a time-effect. There is no time-variable here. They are distributions which don't necessarily have to affect each other. They are nested for the purpose of grouping them. They are compared by looking at the distributions themselves within each sub-group. By looking at the distribution, I am not throwing out any information. Maybe I am missing your point?  In any case, I am wondering how such a Q-Q plot could be made.

[#8]

No worries, I haven't been able to make my point. I use normal plots of effects for crossed studies, but for nested studies I think there are better approaches. I'll move on.

[#9]

@shrirams, I have thought about your situation and the plots you are looking to create.  The diagram you show appears that those components A & B are not nested, but crossed! Every level of A sees every level of B. It seems to me that you are running experiments nested "inside" those components.  This to me sounds like replicates.  The A & B components are either creating 4 blocks (If A & B are noise) or the whole plot (if A & B are design factors) of a split-plot design.    Are these components noise variables or design factors?  In any case, if you are trying to create normal plots for each Block or Whole plot, I have a less than elegant way of doing this.  If my interpretation is correct, I can give you guidance as to how to get JMP to do assist in performing this analysis.

[#10]

@statman Yes indeed, the factors are crossed and not nested. And as you pointed out, I'm looking at the distribution of the replicates. The nesting occurs in the split-plot. In my case, A & B are not noise variables, but rather design factors. Pointers to create such a plot will be very helpful.