Juvenile gun violence dataset: quantitative summary guide

Read the measures instruction again, because it is not offering you a choice you get to make once.

It says to use one of the following — rates, percentages, mean, median or mode — to summarise the results for each of the variables. The parenthesis then names gender, race and age. Those are not the same kind of variable, and that settles which measure is permitted for each. Gender and race are nominal categories with no order and no arithmetic; there is no mean of race, and a median is equally meaningless. Age is a ratio variable and takes a mean or median. So the instruction means one appropriate measure per variable, chosen by the variable's level of measurement — not one measure applied across the dataset. A summary reporting an average race is not a weak answer; it is a statement that cannot be true or false, and a marker teaching quantitative methods will read it as the whole assignment missed.

The word rates in that same list deserves separate attention, because it is the one most likely to be used wrongly and the mistake looks like sophistication. A percentage is a proportion of your sample. A rate is a count divided by the population at risk over a period — offences per 100,000 juveniles, for example. Those denominators are different, and a dataset of juvenile gun offenders contains only offenders. It carries no count of the juveniles who did not offend, so it supplies no denominator, so no rate can be computed from it unless the dataset separately provides population figures. Check before you use the word. If the data cannot support a rate, say percentages and say why, which demonstrates more understanding than producing a number that has no defensible denominator. Naming a limit of your own data is one of the few moves in a quantitative summary that cannot be marked wrong.

The required table is a cross-tabulation of two nominal variables — type of gun by sex of offender — and the decision that determines what it says is one most submissions never make explicitly. You can percentage a contingency table by row or by column, and the two answer different questions. Row percentages, taking each sex as the base, tell you what proportion of male offenders used a handgun and what proportion of female offenders did, which is the comparison the assignment is asking for. Column percentages, taking each gun type as the base, tell you what proportion of handgun users were male — and because juvenile gun offender samples are overwhelmingly male, that version will make every weapon category look male-dominated whether or not the sexes differ at all in what they chose. The table would then report the sex composition of the sample rather than anything about weapon preference, which is easy to miss because the numbers still look meaningful.

So percentage by row, and label the base on the table itself. Then check your cell counts before you interpret anything. Female juvenile gun offenders will be a small subgroup in any dataset of this kind, and a percentage computed on a base of seven moves fourteen points every time one case changes. Report the raw count alongside every percentage rather than percentages alone, and if a difference between the sexes rests on a handful of cases, say so in the sentence where you report it. Nothing damages a quantitative summary faster than a confident claim about a group of nine, and nothing is cheaper to protect against than writing the n in brackets. It also changes how a reader weighs the finding without you having to argue for caution in a separate sentence, which matters when the whole summary has to fit in two pages.

Then comes the part where the assignment asks for something the data cannot give you on its own. Synthesise the results and provide a policy recommendation — but a dataset of offenders describes offenders. It has no comparison group of juveniles who did not offend, so it cannot identify risk factors, cannot establish causes, and cannot tell you what would have happened under a different policy. It can tell you what this group looked like and what they used. A recommendation, meanwhile, is a causal claim: do this and that will change. The gap between those two is real and the assignment does not acknowledge it, so your summary should. Doing that is not a hedge or a refusal of the task; it is the part of the task that demonstrates you understand what kind of data you were given and what conclusions it can carry.

The way to close that gap honestly is to let the data set the target and let external evidence carry the causal claim, then say plainly which is doing which. If the descriptive results show handguns dominating and offenders clustered at the older end of the juvenile range, that identifies where a policy would have to bite. Whether a given policy actually bites is a separate question with its own literature, and there is a standing synthesis that grades exactly this: across the policies reviewed, only four have evidence rated supportive, the strongest category — child-access prevention laws, concealed-carry laws, minimum age requirements and stand-your-ground laws. Everything else sits at moderate, limited or inconclusive, which means most policies a student might reach for have weaker support than the confident register of a recommendation usually implies.

Child-access prevention laws are the ones that fit this population directly, with supportive evidence that they reduce firearm self-injuries including suicides, firearm homicides and assault injuries, and unintentional firearm injuries and deaths among youth. Higher minimum purchase ages have supportive evidence for reducing firearm suicides among young people. Recommending one of those, with its evidence grade named, is a different quality of answer from recommending an education campaign because it sounds constructive. It also lets you be honest about scope: the recommendation is consistent with your descriptive findings and supported by separate causal evidence, and no single claim is asked to do both jobs. That two-part structure is also the easiest thing to mark, because a reader can see immediately which sentence rests on your analysis and which rests on the literature, rather than having to work out where one ends and the other begins.

One piece of context worth a sentence, because it establishes why this analysis matters rather than padding the introduction: firearm injury has become the leading cause of death among children and adolescents aged 1 to 19 in the United States, having surpassed motor vehicle crashes, with firearm deaths in that age group rising 29% between 2019 and 2020. That is a national figure and not a finding from your dataset, so cite it as context and keep it out of the results section — mixing external statistics into your own findings is the other way these summaries lose marks.

Right measure per variable, row percentages with counts shown, and a recommendation whose causal claim is carried by evidence outside the dataset.