That said, if you eat chips, pretzels, candy and such and want to cut down, nuts work great. You can get them bulk online if no stores near you have them. At first they don't have the intense flavors and crunch of typical junk, but the more you eat them the more you find the subtleties in their flavors and textures, much richer and complex than chips and candy. http://joshuaspodek.com/variety-choice-manufactured-illusion
This fall I noticed the nuts I normally got tasted a lot more flavorful. I checked and it turns out nuts are seasonal, so they'll taste different at different times of the year, depending on where you source them. I should have known that, but I enjoyed learning it by noticing the taste. I wouldn't have noticed such subtlety without stopping eating junk food and I find catching those subtle complexities more pleasing than the intensities of junk food.
I throw a bunch on my oatmeal with chopped fruit and chia seeds for a great breakfast that takes less than a minute to prepare. http://joshuaspodek.com/fitness-beliefs-habits-part-2
The mix I get has cashews, almonds, hazelnuts, brazil nuts, and pecans. So much better than pretzels, chips, or candy.
The pecans I don't put in the cereal or eat with the rest. They taste so much better to me. I eat them one at a time so I can savor them specially.
http://www.refsmmat.com/statistics/significant-differences.h...
Skimming through the paper, it appears they did not correct for making multiple comparisons. That will raise their false positive rate. Oh well...
http://www.npr.org/blogs/thesalt/2013/11/21/246549388/nuts-f...
And then my other remark. In the appendix. Table S4: most of the confidence intervals overlap all the way from the 'no nuts'-group to the 'daily nuts'-group. I know you shouldn't focus overly on those intervals. But those splines don't look that convincing to me.
So considering you have a measured effect that is not that clear, where you really expect the "true" effect to be orders smaller than the reported 20% reduction in mortality, without proper controls in the studies for one of the largest effects (higher income >> longer life) and with lots of overlapping confidence intervals, I wouldn't bet my nuts on this study.