How Can We Help?
My correlations are wrong / don’t make sense
When reviewing your correlations, there are a few things to consider here:
1. Improving the quality of your data can improve the reliability of your insights.
- Generally, the more data you have, the more reliable your insights will be. Having at least 7 days with, and 7 days without, a single Factor will help. Always listen to the advice of your health professionals when deciding whether to do more or less of a factor.
- Using time periods in the Symptoms and Other Factors sections can improve the reliability of correlations on the same day. If you don’t use time periods, your correlations the next day may be more reliable.
- Logging changes in your symptoms in the time periods after you’ve logged a Factor can help Bearable know what impact this Factors had. For example, logging your symptom score in the time period (am) before you log a medication (mid) and then also in the time period after you’ve taken that medication (pm).
- Generally, the more data you have, the more reliable your insights will be. Having at least 7 days with, and 7 days without, a single Factor will help. Always listen to the advice of your health professionals when deciding whether to do more or less of a factor.
2. Correlations are not causation.
Bearable can help you to identify patterns between symptoms and factors. However, it cannot tell you whether that pattern is because the factor caused the change in symptoms. Common factors that can affect the quality of a correlation include:
- Reverse relationships, i.e. The factor happens alongside the symptom for other reasons, for example pain medications might be taken on days when you have more pain.
- Confounding variables, i.e. a third variable that influences both the supposed cause and the supposed effect.
- Outliers can distort correlations. Collecting a large enough sample of Symptom and Factor data can help. You can also log significant events (Settings > Significant Events) to help remember circumstances that might be unusual, for example becoming sick.
- Reverse relationships, i.e. The factor happens alongside the symptom for other reasons, for example pain medications might be taken on days when you have more pain.
3. Correlations are open to interpretation.
It’s important to sense check your correlations, if they don’t make sense at first glance, is there anything that might explain this?
For example, you notice that drinking caffeine after 3pm correlates with improved Sleep, which seems counterintuitive, but perhaps there’s a confounding factor such as having a much higher level of activity (i.e. making you more exhausted) on the same days that you typically drink caffeine after 3pm. Using the Grid view in the Impacts tab can help with identifying confounding factors.
If you have a large sample of data, are using time periods, are logging changes in symptoms in the time period following factors i.e. Collecting good quality data, it’s still important to take time to think about what your correlations might mean.
Ultimately, some of the correlations may not be helpful, and it’s just as important for you to identify the correlations that you don’t need to pay attention to. We recommend saving the helpful correlations in the Discoveries tab and making a note of unhelpful correlations in the Extra Notes section to refer back to in the future.
👋 Still have questions?
Let us know by contacting support@bearable.app and a member of the Bearable team can help you with anything you need.
