Normalizing other people's data on number of significant figures

Question

I am presenting a table in my paper which summarizes and compares data from other papers. I'd like the data to be presented uniformly, with the same number of significant figures for all data. However, some of the cited papers have more significant figures in their data than others.

Here's an example: instead of presenting the table below:

    | Smith, J.    | Doe, J.    | Anon, Y.       | My data    |
|---|--------------|------------|----------------|------------|
| X | 1.42+- 0.31  | 1.3 +- 0.1 | 1.314 +- 0.287 | 1.4 +- 0.2 |
| Y | 1.43 +- 0.10 | 1.4 +- 0.2 | 1.421 +- 0.234 | 1.5 +- 0.3 |

I'd like to normalize the number of significant figures to make it easier to compare them:

    | Smith, J.  | Doe, J.    | Anon, Y.   | My data    |
|---|------------|------------|------------|------------|
| X | 1.4 +- 0.3 | 1.3 +- 0.1 | 1.3 +- 0.3 | 1.4 +- 0.2 |
| Y | 1.4 +- 0.1 | 1.4 +- 0.2 | 1.4 +- 0.2 | 1.5 +- 0.3 |

Can I (or should I) remove significant figures from their data, for consistency purposes? If so, can I still use the expression reproduced from Smith, J., or should I use something as adapted from Smith, J.?

Answer 1

You are free to choose how many digits to keep, and whether to do it consistently - although you are advised to use as many as it makes sense.

Reporting 1.314 ± 0.287 clearly does not add any useful information to 1.31 ± 0.29 as there is no meaning in the last digit. As a referee I would have requested the original publication to ignore the last digit.

Keeping Doe, J as 1.3 ± 0.1 together with Anon, Y as 1.31 ± 0.29 in a review makes perfect sense, as it shows that Anon's data is actually more precise than Doe's.

Answer 2

It's certainly a thing you can do, especially if you're simply presenting the data (it might be useful to keep the expanded form to avoid adding any additional rounding error to a meta-analysis or the like). I would simply include a sentence in the Table legend, etc.

All estimates reported to one decimal place for consistency. Or something like that.