inst/ksapp/www/md/operation.md

In geanes/kidstats: Subadult age and sex estimations based on long bone measurements

Operation

Although KidStats can run with R, we recommend using RStudio. RStudio is an integrated development environment for R that works with Shiny, includes Pandoc (a program used to generate reports), and features used to develop the software. RStudio or R needs to be open for KidStats to operate. Thus, if you close R or RStudio the web interface for KidStats will be disabled.

Input

1. Collect the proper measurements (see Measurements section above).
2. Enter these measurements into the appropriate boxes on the Input page.

Age

1. On the left side of the Input page, choose the type of Transformation on age to be performed. A transformation isn't necessary but may provide a better fit, depending on the measurements available. It is acceptable to use all transformation options and choose the fit with the smallest prediction intervals. The results should be fairly similar, no matter what transformation is chosen.
2. Push the Evaluate button to perform the analysis.
3. After analyzing the Output page, you may wish to exclude variables from the modelling process, these variables can be chosen in the Exclude box. The Exclude box is dynamic, so it will only provide measurements to remove from model creation that you have provided measurements for. If you click in the box, the list will appear. Only individuals from the reference sample with all non-excluded measurements entered on the Input page will be used in model creation.
4. Once measurements have been chosen for exclusion from the model, repeat Steps 2 and 3 as necessary.

Age: Output

Upon evaluation, some output is provided directly beneath the measurement input boxes. Specifically, the point estimate, upper and lower 95% prediction intervals, the R-squared value for the model, and the sample size used in the evaluation are provided. A more detailed output is available from the Output page under the Age Estimation tab.

Sex

1. Push the Evaluate button to perform the analysis.
2. After analyzing the Output page (see below), you may wish to exclude variables from the modelling process, these variables can be chosen in the Exclude box. The Exclude box is dynamic, so it will only provide measurements to remove from model creation that you have provided measurements for. If you click in the box, the list will appear.
3. If you wish to calculate the classification accuracy using the bootstrap method, check the Bootstrap Classification Accuracy box.
4. Note: This will slow down the sex estimation process.
5. Once measurements have been chosen for exclusion from the model, repeat Steps 2 and 3 as necessary.

Sex: Output

Some output (male and female posterior probabilities and sample size) is provided directly beneath the measurement input boxes and age estimation results. A more detailed output is available on the Output page under the Sex Estimation tab and includes the posterior probabilities when the unknown is used in the created model, the FDA coefficients, variable importance, a confusion matrix that illustrates the trends in classification/misclassification for the created model, and the bootstrapped classification accuracy.

Comment: It is important, and almost imperative for the sex estimation analyses, to evaluate which variables were incorporated into the model in the extended reports. If variables were not included and have no variable importance (provided by a small evimp number [i.e., 0]) then they can be 'excluded' so they are removed from the next iteration of model creation. The removal of superfluous variables can increase the sample size, which will yield a more realistic prediction interval/classification accuracy. However, remember more variables incorporated into the sex estimation models generally yields higher classification accuracies. The overly optimistic effects of a small sample size can be mitigated by bootstrapping.

geanes/kidstats documentation built on May 17, 2019, 12:15 a.m.