Testing differences between 2 timepoints in the same variable

[Daniel Pozzobon]

Hello everyone!
(This is my first post here, so if I'm doing something wrong, please forgive me - feel fee to correct me)

I'm tryinbg to finish a long delayed Sistematic Review on pain and disability over time.
I have a small dataset consisting of some continuous variables (i.e. pain and disability index) collected across multiple timepoints extracted from some papers included in my search.

I would like to compare the same variable between 2 timepoints (e.g. pain at 3 months vs pain at 6 months) accross multiple studies and have the pvalue of this comparisson, to be able to 'statistically' say "This is equal to that!".
I think I should use ttest to do this, but I cannot -for the life of me- find the correct command structure to run this thing.

If possible, I would like to do the same comparisson but considering longer periods (e.g. pain from 3 to 6 months vs pain from 6 to 12 months).
Is this possible to do this without having my PC (an my brain) melting away?

All suggestions are welcome!
Thank you any and all help provided.


Dataset extert:

id: study id
n: sample size at this timepoint
time: timepoint of data collection since baseline (in months)
mean: mean index (pain/disab) reported

id	n	time	mean
1	92	-6	80
1	92	0	43
1	81	4	68
1	65	12	73
1	51	18	71
2	548	-6	77
2	548	0	27
2	548	4	61
3	35	-6	88
3	35	0,5	53
3	35	3	76
3	31	6	75
3	30	12	84
4	35	0	70
4	35	0,5	53,46
4	35	1,5	40,7
4	35	3	30,56
5	183	0	80
5	183	4	24
5	183	12	69
5	183	18	70
6	61	0	83
6	61	0,5	47
6	61	4	70
6	61	12	74

[George Ford]

I suspect you'd need metan or some such.

Simple way:

g short = inlist(time,0,6)
g longer = inlist(time,7,12)

tabstat mean, by(time)

reg mean short longer 
reg mean short longer [aw=n]

[Daniel Pozzobon]

Thank you very much, Mr. Ford.
I've run your code and I got this output.

Could you (or someone else) help me understand what this galaxy of numbers means?
I have some idea of it, but I dont want to make any assumptions and write things upside-down or backwards on my report.

As per table 1, I get that pain was higher at baseline and got progressively lower as the time went by.
But what does the other 2 tables (Model? Residual? _cons?) means?

Thank you for all your help so far.

[George Ford]

If you are trying to combine results from multiple studies then I suspect the Meta analysis stuff is what you need. I don't do Meta Analysis.

If you are just trying to get a sense of things, then some simple regressions might work.

I don't know what you're after.

It looks to me, with some exceptions, that the pmean is going down over time, except for a bump at 1.5-3 months. But that may be due to some peculiarity in the studies that reported for those times.

[Daniel Pozzobon]

Hello Mr Ford.

I'm trying to check if the diference between timepoints (3 months apart) reaches statistical significance (e.g. pvalue of the diference in pmean from Baseline to 3 months; the same from 3 months vs 6 months; the same from 6 months vs 12 months; and so on).

Thank you very for for all your help.
Your code was very helpfull.