A few final project questions!

A few final project questions!

by Shalini Dixit -

1) For the table, does it have to be a specific format? I know Dr. Pletcher something about the standard format being horizontal lines (no vertical) and I think no color?

2) I also have a question regarding rounding to a different decimal place within the same table. For my table values, I rounded them all to one decimal place (0.1, 0.2, etc.); most of the values represent % of participants. However, for one variable the two percentages are 0.17 and 0.23, which would both end up being 0.2. Is it better to keep them as the original 2-digit values or change them both to 0.2 to be consistent with the rest of the table. The difference between the 0.17 and 0.23 is not statistically significant.

3) Similarly, if the reported p-values have different numbers of decimals, would you just pick the one with the least digits to be consistent? I know questions about conveying p values of "0" have come up before. But just to check, if we have p values of 0.000 and 0.0000, is it preferable to display them all as <0.001 instead of 0.000?

4) For my figure, I did a graph of two Kaplan Meier curves, one for black participants and one for whites, with a failure event of incident atrial fibrillation. For the legend, aside from describing the basic definitions, is it appropriate to include the hazard ratio (and 95% CI and p value) for the adjusted risk of AF among whites, or is that inappropriate because the hazard ratio is not actually displayed in the graph. If not, what statistic would be appropriate or necessary to include?

Thanks for your help and sorry for posting so late!

-Shalini

Re: A few final project questions!

by Michael Peters -

Shalini,

1) Yes, "standard format" is a somewhat relative term, but following the outline Mark put out would be a good template.  Minimize lines where they are not needed, and black and white.  If you need another template feel free to pull a recent NEJM article (hard to go wrong following them).

2) You should choose your rounding based upon each variables precision. You do not need to round all variables to the exact same decimal place.  Just choose to round your digits to something that is reasonable.

3) Yes, display p-values as <0.001 instead of 0.

4) You can report the HR, although it would be even better to report the number at risk for each segment/time interval, again pull up a recent NEJM article to see how they usually report these.  If find this too difficult, throwing the HR would be OK, but you could leave it out all together as your are correct the KM curve does not really display this data.