"Btw, you are using the old version of the merger.R file. The new one does not include the libraries that gave you trouble."

You're right. It's my fault for not having downloaded the 2nd version, because now it finally works.

> write.meta()
Error: could not find function "write.meta"

Perhaps you can open a thread for exposing this basic stuff. I'm no good at R when it comes to prepare data. For example, I don't understand the two # warning sentences here :

read.mega = function(filename){
return(read.csv(filename,sep=";",row.names=1, #this loads the rownames
stringsAsFactors=FALSE, #FACTORS SUCK AVOID LIKE THE PLAGUE
check.names=FALSE)) #avoid prepending X to columns
}

I will try to work with Rcmdr package next time. I know for having tried that I didn't understand what to do with it. I may try Rstudio later. I have noticed that in several youtube videos that users usually work through Rstudio.

The reason why I need menus is because they ensure you're not making mistakes in your coding, and the data window can show you what your columns look like. When you use menus, the output is displayed along with the syntax. That's how I learn to make variables, display stats with the conditional "do if" or perform the analysis repeatedly by groups. See attachment.

I don't understand why they need to make R so complicated. In Stata/SPSS, when you create vars with missing values, you don't need to do stuff like :

use="complete"
na.rm=TRUE

It is not normal that R forces you to learn so much. This is extremely error prone. Of course, the fact that the authors who work with R do not show their syntax, does not make things easier. This is irritating. R is free software, so it is very helpful to show us the syntax.

It is true that R allows weight. But it is not true that you can work with them. For what I know, there is one package you can use to create an object that will weight your analysis.
http://r-survey.r-forge.r-project.org/survey/html/svydesign.html

However, this object works only within the package svydesign.

In other words, you can't do linear regression, correlation, tobit, etc. usually. You have to use the relevant functions within the svydesign. And in this package, I don't see regression, logistic regression, tobit, SEM, factor analysis, etc.

Even if you can create the following object:

dstrat<-svydesign(id=~1,strata=~stype, weights=~pw, data=apistrat, fpc=~fpc)

You cannot do this :

model<-lm(wordsum ~ race + cohort + race:cohort, data=GSSsubset, dstrat)

This is bothersome. It's just like AMOS. In that situation, the only thing you can do with R is to use input data matrix instead of raw data. By using Stata, SAS, or SPSS, you generate your correlation/covariance matrix by using sampling weight. And then switch to R. But in that case, I prefer to do the entire analysis on Stata/SPSS.

-> "Read the help file for read.csv() if you need to know how to read semi-colon separated files."

Ok, but I asked the question because I wanted to work with AMOS. I think SEM should provide graphical output of the results, which AMOS does but not R (at least, not in lavaan package). Furthermore, as I think R is error prone, I prefer to replicate my analysis when using another software.

Concerning the relevant question, the syntax does not display everything I wanted. Notably the partial corr, but I managed to get them by using this :

partial.r(DF.C.PISA.IQ, c(1:18),19)

There are indeed numbers outside the normal range -1;+1, and even a correlation of -2.16. As you know, I hate R for lot of stuff. For example, the syntax above partial out var19, which should be 19th in your list. The name is not displayed, so it's not easy. What is var19, then ? LV2012IQ ?

By the way, why did you use

lm = lm(formula,DF.C.PISA.IQ.Z)

When I once worked with regression, I sometimes standardized my vars before running the regression, but I don't think it's good idea, and anyway, the regression provides both unstandardized and standardized coeff. So, why using z-score then ?

But I didn't understand most of what they say. I gave up many times. I learned new functions but not how to circumvent the many warning messages I got. As I said, R is quite complex. For me, a better presentation would be to compare R with another stats software. I can recommend the following links.

For the write.meta(), I don't know what's happening because I run your syntax from the beginning, including this :

Don't ask me what the error is, I have no idea. And if you don't do insert cbind() into the plot() function, your plot will fail.