While statistical analysis of some parameter shows non significant, it is possible to represent the data in chart or graphical representation
It is recommended to show the region of uncertainty (confidence interval, prediction interval, etc.) around the point estimate in the plots.
If you put side by side on a bar chart diagram, the interval of confidence (either 90, 95 or 99%) then if any overlapping of the confidence interval would mean non-siginficant.
Say group 1 has a mean X and confdence interval X plus or minus 95%CI and a group 2 has mean of Y plus or minus 95%CI. Any overlapping of confidence interval of the two group would indicate no statistical difference between the two groups at level 95%. Same with 90 or 99% vele of confidence. A bar chart with confidence interval or boxplot can add up on the graph confidence interval
see
https://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#boxplot_toc.htm
or proc gchart option vbar with clm=your level of confidence
https://www.sas.com/content/dam/SAS/en_ca/User%20Group%20Presentations/Saskatoon-User-Group/Tagore-Proc-GCHART.pdf
Also this is useful
https://blogs.sas.com/content/iml/2011/10/07/creating-bar-charts-with-confidence-intervals.html
A. Robichaud, it is worth stressing what you mean by giving the different confidence intervals levels and how these map onto the alpha levels of the significance test. I know teaching intro stats some students think that at least in the simple case (with multiple comparisons, for example) that 95% -> 5%. Of course they learn otherwise, but it is worth making this explicit for the RG audience. You would not want readers to think that if the 95% confidence intervals overlap that it implies the test is non-significant at alpha=.05. Attached is an example with p=.01 (I show both the equal variance t-test and Welch's test) and the 95% intervals overlap.
Thanks Daniel and Robichaud.
kindly clarify me
is non-significant data of particular parameter represent in chart ? (Yes/ No)
If yes, How to represent ?
If you show a parameter, like a mean of one group or the slope of a regression, you can show this with a 95% CI and if that doesn't include H0, then it is sig at .05.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis