Recommended way of evaluating RCT with intervention vs. control group for T0 and T1

[maxbri94]

Hi :)

I'm currently evaluating an RCT with two groups (intervention (SCCM) vs. control (TAU)) for two timepoints (T0 = baseline; T1 = post-intervention). I would essentially like to understand the effect of my intervention on the symptom network of my primary outcome measure relative to the control group(depression as measured by PHQ-9). I have constructed four networks, such that:

`# Estimate networks for each model, no tuning

network_model_T0_SCCM <- estimateNetwork(ana_data_T0_SCCM,
default = "EBICglasso",
tuning = 0,
corMethod= "spearman")

network_model_T0_TAU <- estimateNetwork(ana_data_T0_TAU,
default = "EBICglasso",
tuning = 0,
corMethod= "spearman")

network_model_T1_SCCM <- estimateNetwork(ana_data_T1_SCCM,
default = "EBICglasso",
tuning = 0,
corMethod= "spearman")

network_model_T1_TAU <- estimateNetwork(ana_data_T1_TAU,
default = "EBICglasso",
tuning = 0,
corMethod= "spearman")
`

Intuitively, I would make the following four comparisons to understand the treatment effect:

1. network_model_T0_TAU vs. network_model_T0_SCCM --> provides evidence for no baseline differences
2. network_model_T1_TAU vs. network_model_T1_SCCM --> provides evidence for post-intervention differences
3. network_model_T0_TAU vs. network_model_T1_TAU --> provides evidence for changes due to control
4. network_model_T0_SCCM vs. network_model_T1_SCCM --> provides evidence for changes due to treatment

Question 1: The NCT-function in R states that within network comparisons (i.e. 3 & 4 here) have not been validated. Do you recommend against its use?
Question 2: Would it be enough to show that there a no differences in baseline networks but that there are differences in post-intervention networks to ascribe those differences to a treatment effect? I.e. only doing comparisons 1 & 2?

Thank you so much for your help!

Best,

Max