Skip to content

AFNI and NIfTI Server for NIMH/NIH/PHS/DHHS/USA/Earth

Sections
Personal tools
You are here: Home » SSCC » gangc's Home » 1dSEMr: an R program for path analysis

1dSEMr: an R program for path analysis

Document Actions

1dSEMr.R is an R program for model validation in path analysis. For more discussion, see G. Chen, et al., Vector autoregression, structural equation modeling, and their synthesis in neuroimaging data analysis, Comput. Biol. Med. (2011), doi:10.1016/j.compbiomed.2011.09.004 . It comes with the AFNI package, but you can also download it here, and put wherever you want. Start R, and run the program by typing something like

source("~/abin/1dSEMr.R")

and it should be self-evident thereafter. Here is an example verifying the Bullmore et al. (2000) paper with residual variances as input:

source("~/abin/1dSEMr.R")
5
VEC
PFC
SMA
IFG
IPL
1
0.661 1
0.525 0.66 1
0.486 0.507 0.437 1
0.731 0.63 0.558 0.517 1

IPL -> VEC, th1, NA
VEC -> PFC, th2, NA
PFC -> SMA, th3, NA
SMA -> IFG, th4, NA
IFG -> IPL, th5, NA
VEC -> IPL, th6, NA
VEC <-> VEC, NA, 0.825
PFC <-> PFC, NA, 0.868
SMA <-> SMA, NA, 0.87
IFG <-> IFG, NA, 0.881
IPL <-> IPL, NA, 0.851

30

0


We can also analyze the model without providing the residual variances as input:

source("~/abin/1dSEMr.R")
5
VEC
PFC
SMA
IFG
IPL
1
0.661 1
0.525 0.66 1
0.486 0.507 0.437 1
0.731 0.63 0.558 0.517 1

IPL -> VEC, th1, NA
VEC -> PFC, th2, NA
PFC -> SMA, th3, NA
SMA -> IFG, th4, NA
IFG -> IPL, th5, NA
VEC -> IPL, th6, NA
VEC <-> VEC, b1, NA
PFC <-> PFC, b2, NA
SMA <-> SMA, b3, NA
IFG <-> IFG, b4, NA
IPL <-> IPL, b5, NA

30

0


Comparing to the version of 1dSEM


Pros:

* has an interactive mode (can be set in a batch mode too);
* allows the user to specify residuals as unknown parameters in the model;
* can run SEM at individual or group level
* can constrain two or more paths with equal strength;
* provides significance testing for each path;
* provides standard error for each path so that confidence interval can be obtained;
* provides more fit indexes.

Cons:

* there is no model search option;
* estimating degrees of freedom is moot. Since there is intrinsic temporal correlation in the FMRI time series, we can't simply use the number of time points as degrees of freedom, and I don't have an easy solution right now.

(not specified)
Last modified 2011-10-08 17:32
 

Powered by Plone

This site conforms to the following standards: