Long table and wide table

[li zhi]

Hi , all！
If I have two sheets ，sheet1 shows every students‘ score got in the exam and sheet2 contains their physical characteristics.
sheet1

name	gender	year	Math	English	Science
Tom	m	2018	66	93	89
Tom	m	2019	79	100	74
Tom	m	2020	74	84	85
Jerry	m	2018	78	64	63
Jerry	m	2019	77	64	98
Jerry	m	2020	96	60	94
Alice	f	2018	97	85	100
Alice	f	2019	66	70	100
Alice	f	2020	65	63	94

sheet2

name	gender	year	height	weight
Tom	m	2019	178	75
Tom	m	2020	178	78
Jerry	m	2019	173	70
Jerry	m	2020	173	72
Alice	f	2019	165	58
Alice	f	2020	165	60

Should I reshape them to wide table and merge them to one sheet like below ? I can easily see 3 observations here and directly study the relationship between scores and body features without calling different dtas saved in different folders, but it makes variable's name more complex. Is there any rule
about when to use long table and when to use wide table? Any suggestions ？ Thanks in advance.

name	gender	Math_2018	Math_2019	Math_2020	English_2018	English_2019	English_2020	Science_2018	Science_2019	Science_2020	height_2018	height_2019	weight_2018	weight_2019
Tom	m	66	79	74	93	100	84	89	74	85	178	178	75	78
Jerry	m	78	77	96	64	64	60	63	98	94	173	173	70	72
Alice	f	97	66	65	85	70	63	100	100	94	165	165	58	60

[Joro Kolev]

I am not entirely sure what you are asking.

You can equally well merge if your datasets are in wide, and in long format. Merging does not become any easier (or harder) if you make them wide. Hence merging is not a reason to prefer one over the other.

The main reason to prefer wide over long or the other way round is the model you are going to fit to your data.

For example if I regress the test score in particular year on the weight of the student in this and in the other years, then I would have to use the wide form.

If I regress score on weight but only in the particular year, then long would do.

In short, long vs wide is mainly determined by the analysis you want to carry out.

[li zhi]

I didn't give a good example.

I mean when there are many sheets and all of them are complex , for example, the values of some variable in each sheet are not unique, the command "merge m:m " may cause mistakes.

I already have the separate tables , and I want to put them together , but I am not sure if this action is reasonable.

Maybe I should just keep them in separate and call the needed data when running the models.

So I want to learn about a general way to deal with this situation.

[Nick Cox]

#2

For example if I regress the test score in particular year on the weight of the student in this and in the other years, then I would have to use the wide form.

Not so, as you can use time series operators directly. Long is fine for that kind of problem.