Overlapping and non-overlapping error bars may indicate non-significant and significant differences between respective groups. As mentioned today, it becomes difficult to resolve significant differences between groups if your error bars don’t clearly visually overlap. Hence, the t-test is done to compare between groups to definitively make you confident about describing the trend/patterns you see (i.e. reliability of data).
Confidence intervals are one way of discussing the reliability of data but not the best way. It is certainly NOT used in practice (professional science).
Dr G’s Statistics Survival Kit (Townsville 2019)
Dr Gurion Ang from the School of Biological Sciences, The University of Queensland, will visit Townsville 23-25 July 2019. He will deliver statistics workshops to Year 11 Biology and Marine Science students at various schools in Townsville. Twilight Teacher CPD sessions will be available at no cost. Find out more and register here: bit.ly/ssktownsville