0

I attempted to perform t-sne on some variables: Probability (%) of dying between age 30 and exact age 70 from illness (all)', 'Suicides per 100000 (all)', 'social_support', 'birth_health', 'freedom', 'generosity', 'perception_corruption', 'government_confidence', 'percent_internet_users', 'GDP_per_person', 'gini_reported', 'temperature_change'.

You see the scattered result below. I set the perplexity parameter to 3. I want to investigate how the different variables impact a countries happiness. As far I know there is no direct way to interpret the meaning of the x axis and y axis. Also there seem to be no direct clusters. That's a bad sign, right? Changing the perplexity parameter does not change the general structure a lot. Along the negative direction of the x-axis the happiness seems to increase. The y-axis seems to have not a significant impact. Can you help me interpret this?

tsne

thingy
  • 1
  • Presumably the lack of clusters indicates that separation may be difficult. You might look to seek whether there are geographical groups in your chart, such as South America or western Europe or Gulf states. Alternatively, you could try looking at PCA-based charts (so more easily interpreted) and whether any of those seem associated with happiness – Henry Sep 06 '23 at 16:57
  • Hi, usually it is not so easy and safe to draw hard conclusions from t-SNE. Here is an excellent post on u-maps and t-SNE which might help you out: https://pair-code.github.io/understanding-umap/ – Janosch Sep 06 '23 at 18:04
  • My point #3 here might be of interest. – Dave Sep 06 '23 at 18:27

0 Answers0