Does Statistica support Nested Gage R&R for destructive testing?
Article ID: KB0075804
Updated On:
| Products | Versions |
|---|---|
| Spotfire Statistica | 13.0 and higher |
Description
In Statistica, the Gage R&R In Statistica, the Gage R&R function under "Process Analysis" is for Crossed effects only (operator by part interaction). Each Part is measured repeatedly by each Operator.
Does Statistica support Nested Gage R&R for destructive testing?
Does Statistica support Nested Gage R&R for destructive testing?
Issue/Introduction
Does Statistica support Nested Gage R&R for destructive testing?
Environment
Windows
Resolution
Nested Gage R&R for destructive testing mimics a simple nested ANOVA with random factors. A user could analyze this within Statistica GLM or Variance Estimation and Precision modules.
The statistical problem addressed in Gage R&R studies is variance estimation. The main/essential difference between a "regular" ANOVA and Gage R&R is that operators and parts are treated as random effects.
In order to estimate the variances, the effects need to be treated or specified as random within the software.
For illustration, use Statistica example dataset "temperat.sta",
1. In Statistica menu "Statistics > Advanced Models > General Linear", you can choose "Analysis wizard', select "Measure" as dependent and "Operator" and "Part" as Grouping variables, under "Custom between design", specify the model design terms "Operator", "Part", "Operator * Part", click "Next", and specify them as "Random Factors", run the analysis.
In the result dialog, under "summary" tab, click on "Var. comps" under "Random effects" to display the variance components.
2. Other than GLM module, you can alternatively use Statistica "Statistics > Variance" (Variance Estimation and Precision) to specify the terms as Random effects (Operator, Part, Operator * Part) under "Customize Design" option:
3. Also, variance components are additive, so to get percentages you just need to compute ratios of the variance component over the total variance.
For example, the percentage for operator is given as the following:
Var( Operator ) / [ Var(Operator) + Var(Parts) + Var(Op x Parts) + Var(Error) ]
Note:
You can also refer to Statistica e-manual example under section "Variance Estimation and Precision": "Contents > Statistics and Analyzing data > Statistics > Variance Estimation and Precision > Examples > Example 3: Variance Component Estimation for a Hierarchically Nested Random Design":
The statistical problem addressed in Gage R&R studies is variance estimation. The main/essential difference between a "regular" ANOVA and Gage R&R is that operators and parts are treated as random effects.
In order to estimate the variances, the effects need to be treated or specified as random within the software.
For illustration, use Statistica example dataset "temperat.sta",
1. In Statistica menu "Statistics > Advanced Models > General Linear", you can choose "Analysis wizard', select "Measure" as dependent and "Operator" and "Part" as Grouping variables, under "Custom between design", specify the model design terms "Operator", "Part", "Operator * Part", click "Next", and specify them as "Random Factors", run the analysis.
In the result dialog, under "summary" tab, click on "Var. comps" under "Random effects" to display the variance components.
2. Other than GLM module, you can alternatively use Statistica "Statistics > Variance" (Variance Estimation and Precision) to specify the terms as Random effects (Operator, Part, Operator * Part) under "Customize Design" option:
3. Also, variance components are additive, so to get percentages you just need to compute ratios of the variance component over the total variance.
For example, the percentage for operator is given as the following:
Var( Operator ) / [ Var(Operator) + Var(Parts) + Var(Op x Parts) + Var(Error) ]
Note:
You can also refer to Statistica e-manual example under section "Variance Estimation and Precision": "Contents > Statistics and Analyzing data > Statistics > Variance Estimation and Precision > Examples > Example 3: Variance Component Estimation for a Hierarchically Nested Random Design":
Feedback
Yes
No
Powered by
