Continuous independent variable

[dxkeac]

Hi, thanks for a nice tool! I have a question regarding the usage of FLASHMM. I am working with single-cell data from 10 samples, each measured at three time points. I would like to model the temporal effect by treating the time points as a continuous variable (e.g., 0, 1, 2), rather than as categorical factors.

Would it be appropriate to use FLASHMM in this setting to model continuous time effects? Any guidance or recommended practices would be greatly appreciated. Thank you!

[ChangjiangXu]

You may use FLASHMM by setting the time as a covariate and the sample as a random factor. For example, define the fixed and random effects formulas as fixed = ~ time + other covariates and random = ~ 0 + sample, and run flashmm function in the FLASHMM. Or let X = model.matrix(fixed) and Z = model.matrix(random), and run lmmfit function.

[dxkeac]

Thank you very much! I have one quick follow-up question: I also have bulk RNA-seq data, would FLASHMM still be appropriate for modeling longitudinal bulk RNA-seq data with repeated measures, each including three time points?
Thank you again for your time and help!

[ChangjiangXu]

Yes, FLASHMM works for the longitudinal bulk RNA-seq data. For the bulk RNA-seq data, you may also use lme4 package.

[dxkeac]

Thank you very much for the suggestion. I will try FLASHMM, since I would like to use a faster method.

[dxkeac]

I am using lmmfit for longitudinal bulk RNA-seq data with the following model:

fixed <- ~ time_c + time2_c + PC1 + PC3
random <- ~ 0 + subject
#fit <- lmmfit(Y, X, Z, d = ncol(Z))
The result came out very quickly!

I would like to perform a joint hypothesis test for the temporal effects, specifically:

H0: β_time_c = 0 and β_time2_c = 0

I am wondering if lmmtest() supports specifying a vector that jointly tests multiple fixed effects (e.g., time_c and time2_c)?

Thank you!

[ChangjiangXu]

The lmmtest function doesn't test if a vector of parameters is zero. In this case, likelihood ratio test (LRT) can be used.