Error in Factor Analysis - Starting Values

Question

I have a n = 1008 x p = 45 data matrix and I am trying to perform a factor analysis in the data, using the factanal() function. I am getting the same error regardless of the number of factors that I decide to fit: "Error in factanal(rios, 3, rotation = "varimax") : unable to optimize from this starting value"

I have tried several sets of uniqueness starting values, using the start argument, and even so, none seems to fit. Can someone help me?

My data consists of monthly outflows (in m³/s) from 45 Brazilian rivers (84 years x 12 months = 1008 observations). The rivers are the variables (p = 45). I have attached a link to my .csv file containing such data:

https://www.dropbox.com/s/swz1mqxmol7tofx/Rivers%20as%20variables%20%28csv%29.csv?dl=0

Thanks in advance!

Coakley

Answer 1

Try increasing the lower bound for uniquenesses during optimization, so it allows a solution to converge.

Increasing the lower bound from the default 0.005 to 0.1 works in this case.

factanal(rios, 3, rotation = "varimax", lower = 0.01) 

Refer to the documentation of facanal for more details.

Blau

Answer 2

Try with multiple start values. This worked in my case:

factanal(rios, 3, nstart = 4)

Preparedness