Update Scenario1.md

docs/Scenario1.md

@@ -387,6 +387,8 @@ You can also assign your own weights to the two different classes of observation
 
 (iii) **Calculate partial dependencies** Partial dependencies is another *global explainability* method, that helps you understand the relationship between individual feature values and classification probabilities. For an accessable description of partial dependencies, see [this blog post](https://towardsdatascience.com/partial-dependence-plots-with-scikit-learn-966ace4864fc). Ticking this box will calculate partial dependencies for ecery feature in your data set, and one CSV file will be saved for every feature. For an example of this CSV file, click [HERE]. BEWARE: Calculating parial dependencies come with long run-times.
 
+(iv) **Create train/test video and frame index log** When training a model, SimBA takes all your data inside your `project_folder/csv/targets_inserted` directory, and splits it into a _training_ set, and a _testing_ set. How those two sets look like, largely depend on your defined sampling settings above. Let's say we have trained our model, and we want to check which frames, in which videos, was actually used to train and test the model. If this checkbox is ticked, two files will be produced in models foler subdirectories: [train_idx.csv](https://github.com/sgoldenlab/simba/blob/master/misc/train_idx_example.csv) and [test_idx.csv](https://github.com/sgoldenlab/simba/blob/master/misc/test_idx_example.csv), that lists the frame numbers from the repective videos used in the training and testing of your models(s).
+
 #### Train predictive classifier(s): start the machine training
 
 Once all the entry boxes have been filled in with the desired Hyperparameters and Model Evaluation Settings, the user can either click on `Save settings into global environment` or `Save settings for specific model`. If you click on `Save settings into global environment`, the settings will be saved in to your *project_config.ini* file located in your `project_folder`. These settings can subsequently be retreived and executed to generate a predictive classifier (**Mode 1**).