top of page

Help Centre

For each of the tabs in COMPLEX-IT we provide tutorial videos, PDFs for reading, research articles, and (where useful) additional links to explore.

THESE TUTORIALS ARE FOR BETA 1.0.0

Tutorials for our Current Beta 1.0.1. are forthcoming;

in the meantime, please use those below as general guides. 

Tutorial 1: Introduction to Case-Based Complexity
 

The purpose of this tutorial is to introduce readers to the methodology of Case-Based Complexity.

Case-based complexity is the study of cases and their comparison, and their underlying complex causality, in complex systems terms; which scholars do by employing various methodological combinations of the complexity sciences, complex realism, qualitative-comparative analysis, systems mapping and computational modelling.

Screenshot 2023-09-12 at 13.09.00.png

Click here to explore a repository of articles on case-based complexity and case-based modelling

Tutorial 2: Introduction to COMPLEX-IT
 

The purpose of this tutorial is to introduce readers to COMPLEX-IT and the new field of AM-Smart methods.

COMPLEX-IT is a case-based, mixed-methods platform for applied social inquiry to complex data/systems, designed to increase non-expert access to the tools of computational social science. Presently, the platform is comprised of cluster analysis, artificial intelligence, data visualization, data forecasting, case-based systems mapping, and case-based scenario simulation.

 

COMPLEX-IT is part of the new methods field called approachable modeling and smart methods -- or AM-Smart for short -- designed to increase non-expert access to computational modelling, as well as help them to learn these methods.

Screenshot 2023-09-12 at 13.44.37.png
Screenshot 2023-09-12 at 13.47.46.png
Tutorial 3: How to import your database in Tab 1

The purpose of this tutorial is to show you how to upload, inspect and manage your dataset.

  • Your data must be in the form of a csv file.

  • For more on creating csv files CLICK HERE

  • NOTE 1: The online version erases your data after each session. So you have to upload your data each time you use the online version.

  • NOTE 2: Depending upon the software you use, your csv files may be saved with commas or other separators. COMPLEX-IT allows you to select for different forms.

  • SUB-SET YOUR DATA: During your analyses, you may want to subset your data to explore a particular group of factors. You can do so by returning to TAB1, turning off the factors you do not want to explore and then clicking the SUBSET. 

tutorial-data.jpg

PRACTICE DATASET: To help you practice we have provided an example dataset from the Wales Indices of Multiple Deprivation.

  • The EXCEL file explains the dataset.

  • The CSV file is the one you actually upload.

  • See also the FILE SHARE section of this website for other shared datasets.

CSVtab.jpg
EXCELtab.jpg
Tutorial 4: How to run cluster analysis in Tab 2

The purpose of this tutorial is to show you how to run, explore and confirm the cluster solution for your data

  • As you conduct your analyses, all of the statistical results are stored in the downloadable report in Tab 8. So no need to worry about capturing that data here.

  • The goal here is to explore various cluster solutions in a visually intuitive environment that does not require expertise in cluster analysis, per say

STEP 1: HYPOTHESIZE YOUR CLUSTER SOLUTION

  1. To begin, how many clusters do you think are in your database?

  2. What is your hypothesis based on -- the literature, a guess, expertise, experience, a hunch?

  3. How would you describe or name these different clusters?

  4. How do you think your case-based profile of variables account for these different clusters?

 

STEP 2: RUN THE K-MEANS

  1. Run your k-means several times to see if you can improve the Pseudo F

  2. How strong is the Pseudo F for your solution?

  3. Looking at the Silhouette, how well are the cases distributed for each cluster?

  4. Should you re-run k-means to look for more or less clsuters?

  5. SOME DEFINITIONS: The Pseudo F indicates the quality of the overall solution; the larger the number, the better the fit.

  6. The Silhouette displays how well each case fits within its respective cluster; where a score of 1 is a perfect fit.

  7. NOTE: The K-means solution and related statistics are found in the GENERATE REPORT TAB.

tutorial-cluster.jpg
Tutorial 5: How to use AI to confirm your cluster solution

The purpose of this tutorial is to show you how to run the AI tab to confirm your cluster solution

  • Just like Tab 2, as you conduct your analyses, all of the statistical results are stored in the downloadable report in Tab 8. So no need to worry about capturing that data here.

TRAINING YOUR AI (Self-Organizing Neural Net)

  1. For those new to the SOM, we recommend using all of the defaults. Beyond that, there is not much to it. Just hit TRAIN SOM.

  2. For those seeking to make use of the options, we recommend watching the tutorial video or reading the tutorial PDF and related articles.

  3. For those running the desktop version and want to explore how to modify or run additional analyses, click on the cran.r-project.org link.

INTERPRETING YOUR AI RESULTS

  1. Results come in two forms:

    1. The degree of fit, provided as two forms of error.

      1. Topographical and Quantitative.

      2. The closer to zero the better.

    2. Importance of each factor, including its significance, to the solution.

      1. Like an F score, the larger the value, the more important the factor. Significance = asterisk.

  2. NOTE: The Report in Tab 8 saves your AI solution for future usage or for running in other packages.

tutorial-SOM.jpg
Tutorial 6: How to visually explore your cluster solutions

The purpose of this tutorial is to show you how to use the data visualization tab to explore your cluster solution and the AI cluster solution.

  • As with previous tabs, as you conduct your analyses, all of the statistical results are stored in the downloadable report in Tab 8. So no need to worry about capturing that data here.

Screenshot 2023-12-09 at 16.23.18.png

READING THE SOM GRID:

  1. To begin, we label each case with its CASE ID and K-MEANS ID.

  2. To see these IDs, for 'PLOT WHAT?' select observations; and for 'TYPE OF PLOT' select names.

  3. The first ID on the grid is the k-means cluster number; the second ID is the case.

  4. The grid also places each case in a quadrant, based on the SOM AI solution.

  5. The more similar the profile, the closer the cases on the grid; the more profiles differ, the further away cases are.

  6. The PROTOTYPES option (i.e., variables) shows how your profile of variables influenced where cases are located.

  7. The BARPLOT option for both OBSERVATIONS and PROTOTYPES shows the profile of variables for each quadrant.

  8. The line in the BARPLOT is mean=0; above the line is more of a variable;below the line is less.

  9. NOTE: Several of the images created here are found in the GENERATE REPORT TAB.

  10. In addition, we recommend using SCREEN CAPTURE to save an image

INTERPRETING YOUR RESULTS:

  1. Looking at the Names, are cases with similar k-means IDs located in similar quadrants?

  2. If yes, do you think the SOM and k-means are reasonably similar solutions? Or, should you re=run your k-means?

  3. How do the profiles account for the different cluster solutions and the quadrant locations of the cases?

  4. What factors (i.e., variables) seem to have the biggest impact on different clusters or the model as a whole?

  5. How does the data solution differ from your hypotheses back at the design phase of COMPLEX-IT?

  6. Are you satisfied with your solution? If not, go back and run your k-means and SOM again.

Tutorial 7: Rerunning your Cluster Analysis and SOM
 

The purpose of this tutorial is to introduce readers to the data mining idea of re-running your cluster analysis and SOM solution as you learn new things about your data or wish to explore other aspects.

Tutorial 8: Using your model to run the scenario simulation tab

The purpose of this tutorial is to introduce the scenario simulation tab.

Here we will use your model to explore different scenarios, policies, and interventions. To do that, we will be using your k-means clusters and your SOM AI solution and grid

 

The grid in this tab visually displays the results of your k-means and SOM AI cluster solutions.

Screenshot 2023-12-09 at 16.35.55.png

TO RUN MODEL:

  1. Start by clicking on MODEL SETUP, which creates the SOM grid created with TAB4

  2. The grid you see is based on the SOM solution you arrived at using TAB3

  3. Next, click the RUN CLUSTERS tab, which places your k-means solution on the SOM grid

  4. These are the k-means clusters you settled on using TAB2

  5. Next, make changes to the various profile of variables for each of the cases.

  6. Once done, click on RUN CLUSTERS again, to see if and where on the grid the cluster moved

  7. Next, look at the BARPLOT grid to see what profile of factors account for the new grid placement

  8. Is this where you wanted your cluster to arrive? If not, try changing something else

  9. If satisfied with your solution, run SENSITIVITY ANALYSIS; if not, click MODEL SETUP to reset.

 

TO RUN SENSITIVITY ANALYSES:

  1. Pick the CLUSTER you are testing from the options

  2. Decide how much to dither your solution by in order to account for variance and error that go with any real-world estimation of change

  3. Run the sensitivity analysis

  4. NOTE: very complex solutions can several minutes or hours to finish

Tutorial 9: Running the Prediction/Forecasting Tab
 

The purpose of this tutorial is to explore how to use the data prediction and forecasting tab.

RUNNING THE PREDICTION TAB:

  1. To begin, you need to convert your new dataset into a CSV file

  2. This CSV file can be comprised of a single new case or a large dataset of new cases

  3. Decide to use the SOM solution from your current session or a previously saved SOM solution

  4. Click CLASSIFY PROFILES

  5. NOTE: you can find your results saved in the GENERATE REPORT TAB

  6. The programme will crash if the headers/format of your new dataset are not the same as the TAB 1 dataset

 

INTERPRETING YOUR RESULTS:

  1. After you run the data, you get a list of each case

  2. For each case, you will see its variable profile

  3. This is followed by the SOM grid quadrant that best fits it

  4. For validity purposes, COMPLEX-IT also provides the second best grid quadrant fit

  5. NOTE: For advanced users, goodness-of-fit for classification is based on a numeric tolerance defined as 10^(-10)

Tutorial 10: Running the Systems mapping tab
 

The purpose of this tutorial is to explore how to use the systems mapping tab.

LEARNING HOW TO THINKING ABOUT THE SYSTEMS MAP:

  1. The tab is intended to help you visually think about the relationships amongst your variables as a network of connections and pathways of influence

  2. It shows the correlation of pairs of factor, and encourages you to evaluate them, add new nodes and connections which represent your beliefs about possible causal connections, or pull out subsection of the map.

  3. THINK, DON’T ACCEPT: Be careful not to interpret the map as causal connections. It is only showing you the correlation between pairs of nodes.

  4. Keep in mind the correlations are not conditioned on other factors (i.e. they do not control for other variables), so we must keep a critical mindset – these maps are intended to prompt thinking and discussion, not offer definitive or ‘correct’ analysis.

 

USING THE SYSTEM MAP TO EXPLORE YOUR DATA:

  1. Along the left you will see various toggles to change your network. Changing these makes 'deep' changes to the network, as it influences the final data frame informing the construction of the network. This means changes here can be combined and carried over between changes to these toggles.

  2. The network itself is rendered using the visNetwork package. Using this you can add and remove nodes and edges, and change the position of nodes. Be aware: these are 'shallow' or aesthetic changes: changing any parameter on the left will erase any changes made.

  3. Along the bottom you can examine node and network statistic information.

Screenshot 2023-12-09 at 17.19.40.png
Screenshot 2023-12-09 at 17.23.59.png
Screenshot 2023-12-09 at 17.28.42.png
Tutorial 11: Generating your report and exploring results
 

The purpose of this tutorial is to introduce readers to how to generate the final report from all of your analyses and to then explore the various outputs, including EXCEL sheets and .jpegs

bottom of page