One of the comments came from a colleague of mine (@katpyxa) and I would like to dedicate this post to addressing it.
Reading previous posts you probably realized I am tend to remove the grid lines in the background as a "non-data ink" and something that clutters the graph driving attention away from the data. It was pointed out that it may not always be a good solution. Indeed, if you want you readers to be able to read the exact values from the graph keeping grid lines is very useful.
For example (taken from Li et al. 2017):
redone to:
Here, I applied some rules from previous posts, namely:
1) depicted values as lengths rather than heights (notice how much easier it is to read the labels)
2) made it right in black and white (in the original paper color was linked to another figure, so this works for the purpose of this post only)
3) added the grid lines. Importantly, grid lines are lighter shade of grey than the bars, indicating it is not as important as other parts of the graph (non-data), at the same time supporting the readability of the values.
Other time when it makes sense to keep grids, is if your values are normalized to a control and you want to emphasize the control value, like here (Martorell-Riera et al. 2014):
Which can be redone to:
Alternatively, the overall graph can be separated per condition, as follows:
Overall, if you like the grid keep it, make sure though it acts as support for reading your data and carries less weight than actual data.
On the final note, I would like to further encourage you to give me comments, send me ideas for makeovers or issues you would like addressed when it comes to your figures.
