Scatter graph plot (time series) with multiple observations

[joseph dover]

Hi there,

I am looking at marks of 200 students over time.

I have a scatter graph with:
TIME: X-axis
MARKS: Y-axis

"twoway (scatter MARKS TIME, sort)"

I have 200 observations, hence my scatter graph is a bit messy

How could I plot an line of the average marks of the class? (instead of having plenty of dots everywhere?)

PS: The command : "twoway (line MARKS TIME, sort)" does not work as it connects all the dots of every students (and there are 200 dots for every year)

Regards,

Joseph

[wbuchanan]

Is the unit of measurement students or classes? If classes you'd want to collapse the data into class level means first. If you wanted to help people understand the relationship being displayed by the scatterplot you could use:

tw scatter MARKS TIME, sort || lfit MARKS TIME

To overlay the regression of marks on time over the scatterplot points.

[Nick Cox]

There are many kinds of averaging. See also

* a variant of William Buchanan's approach 
ssc desc aaplot  

ssc desc rcspline 

ssc desc localp 

help lpoly 

search polarsm

[joseph dover]

Hi wbuchanan!

My unit of measurement is by students (thus 200 students = 200 observations per test)
And I would like to obtain a line which will be the average of the class over time (200 students in the class)

Thanks

Joseph

[Nick Cox]

Further, if you just want a graph of averages for a few distinct times, you could go

egen mean = mean(marks), by(time
scatter marks time || connected mean time, sort

although localp (SSC) or lpoly may well give you similar graphs with suitable bandwith choice. That depends on whether time is clustered or not.

http://www.statalist.org/forums/foru...thodox-methods makes this vivid.

[joseph dover]

It worked! Thank you for your help