Time series for multiple subjects

Home Forums Technical support Time series for multiple subjects

This topic contains 2 replies, has 2 voices, and was last updated by  Vaughan 1 month, 2 weeks ago.

Viewing 3 posts - 1 through 3 (of 3 total)
  • Author
  • #804 Reply


    Thanks to Sacha’s excellent online cookbook (many thanks!) I’ve been attempting time series modelling for the first time.

    We have PSYRATS assessment for patients, conducted at 6 time points with increasing dropouts over time (it’s clinical assessment data) and I’m keen to look at changes over time.

    However, mlVar seems to want very large samples. For example, here’s the minimum it seems to like when simulating data to not spit out an error.

    Model <- mlVARsim(nPerson = 800, nTime = 6, nNode = 3)
    Data <- Model$Data
    Res <- mlVAR(Data,vars = c(“V1″,”V2″,”V3″), idvar = “ID”)

    We’ve got more like N=400 at t1, N=300 at t2 etc (and we’d be happy to look at just two time points in the first instance) with either 6 or 11 items, which seems impossible.

    Am I right in thinking that this just reflects the minimum amount of data needed for a time series analysis or might there be another way to approach the problem?

    Many thanks,

    #806 Reply

    Sacha Epskamp

    Hi Vaughan,

    This mainly have to do with the number of time points. 6 time points is very very little. I’d suggest then only estimating fixed effects, e.g., using graphicalVAR.

    Best, Sacha

    #807 Reply


    Really useful, thanks Sacha. I’ll try and see what happens.

Viewing 3 posts - 1 through 3 (of 3 total)
Reply To: Time series for multiple subjects
Your information: