[1] http://www.ncbi.nlm.nih.gov/pubmed/16542786
[2] http://www.ncbi.nlm.nih.gov/pubmed/19944540
[3] https://examine.com/supplements/magnesium/#summary3-7
[4] https://www.amazon.com/KAL-Magnesium-Glycinate-400-tablets/d...
[1] http://www.ncbi.nlm.nih.gov/pubmed/16542786
[2] http://www.ncbi.nlm.nih.gov/pubmed/19944540
[3] https://examine.com/supplements/magnesium/#summary3-7
[4] https://www.amazon.com/KAL-Magnesium-Glycinate-400-tablets/d...
[1] Refers to case studies (e.g. clinical anecdotes) of depression recovery following magnesium administration, but the number of cases is omitted. More importantly, there are no placebo controls to differentiate between spontaneous or placebo response and response to the magnesium supplementation. Finally, the article is in the context of severe magnesium deficiency, which is a strong selection bias that prevents the anecdotes from being extrapolated to the general population anyway. (Yes, I am aware that much of the US population does not meet FDA guidelines for magnesium intake, but that's not the same as being deficient in magnesium in the body).
[2] Is a good example of the importance of noting which journal the article was published in: In this case, it was published in Medical Hypotheses, which is dedicated to promoting diversity in debate by publishing hypothesis and fragmentary ideas that don't yet have enough real evidence to publish elsewhere. The first study it cites is from 1921, which was not placebo-controlled. The second study cited is from a journal dedicated to Magnesium research, but more importantly which pre-selected for patients with hypomagnesia. In other words, the study examined whether Imipramine (an anti-depressant) or magnesium supplementation were more effective in elderly type 2 diabetes patients who were known to have hypomagnesia.
[3] Refers to animal studies, but 3 of the 4 cited articles are about magnesium in migraine headaches. Perhaps the full-text articles have some embedded murine studies, but otherwise I don't see much evidence for magnesium treatment in humans here.
In short: It can't hurt to supplement, but it's unlikely to be curative. More importantly, recommending weak or unproven supplements to depressed patients can send them down a path of attempted self-medication with supplements that do not work, when proper medical treatment can work much more quickly and effectively.