PHILADELPHIA —Meleanie Hain of Lebanon, Pa., used to tell the news media that she carried a Glock 26 pistol everywhere she went to protect herself and her children. Then last week she was shot to death by her husband in what police called a murder-suicide.

For years, researchers have been trying to investigate whether carrying a gun is protective or risky. But getting the answer through science has proved elusive. Now, University of Pennsylvania researcher Charles Branas has tried a new tack — employing methods normally used by epidemiologists to study cancer and other diseases.

Branas compared a group of shooting victims to a similar set of “controls” who had not been shot. His results, he said, show that guns did not, on average, protect the owners from being shot in an assault — and in fact raised the risk by four times or more.

The study was published in the current issue of the American Journal of Public Health.

Several statisticians, however, called this conclusion a stretch.

Where the experts do agree is on the need for solid scientific information about the risks or benefits of guns.

But gun research is fraught with difficulty, the experts say. Not only is it politically and emotionally charged, but privacy issues make it hard to get large-scale data on who owns a gun and who carries one.

Branas, a former paramedic trained as an epidemiologist, designed an ambitious study that he said stemmed from his experience transporting victims of urban violence.

For this study, he and his colleagues relied on police to get information on shootings in Philadelphia between 2003 and 2006 — a total of 3,485.

The researchers got the locations, victims’ description and whether they had guns with them at the time.

Researchers randomly chose 677 of the victims for the study, from various occupations — taxi drivers, bartenders, nurses and drug dealers. Fifty-three percent had criminal records. Six percent had guns with them when they were shot.

Branas then compared this “case” group with a group of “control” subjects — similar residents of Philadelphia who had not been shot. Controls were matched to each victim according to race, age and sex. The controls were called soon after each shooting and asked whether they had a gun close by during the same 15- to 30-minute interval. The goal was to see if those who got shot were likelier to have guns.

The controls were equally likely to have a gun with them, but more than 80 percent of them were at home at the time of the incident, and many more people own guns than carry them. Only 9 percent of the victims were home when they were shot.

Then Branas said he made statistical corrections for this and other factors that might influence a person’s chance of being a victim, such as neighborhood type, a person’s use of alcohol, and involvement in the drug trade.

After all the corrections were put in, he and his colleagues concluded that the people in the study who were carrying a gun at any given time interval were more than four times as likely to be shot.

Branas offered several possible explanations. Having guns could induce people to behave differently, he said, perhaps emboldening them. Another possibility, he said, is that people are having their firearms turned on them.

But several statisticians not involved in the gun research said you can’t reach such sweeping conclusions with this kind of study.

Criminologist Gary Kleck of Florida State University said the Penn results can be explained by the fact that people who are at risk of being shot are also more likely to buy or carry guns. Such people might have dangerous jobs or belong to a street gang or be involved in the illicit drug trade, for example.

In an e-mail, Kleck explained his view with an analogy. “It is precisely as if medical researchers found that insulin use is more common among persons who suffer from diabetes than among those who are not diabetic (something that is most assuredly true), and concluded that insulin use raises one’s risk of diabetes.”