Login

News

How are weights used in mxData

Posted on
No user picture. ZR Joined: 07/26/2017
Forums
General SEM Discussions
Hi,

I couldn't find a in depth enough explanation in the documentation and I don't know c++ to be able to get this information from the source code, but I just wondered if anyone can expalin exactly how the weights in the "weights" argument in mxData is actually used mathematically? I know obviously a bigger weight will give more weight to that row of data etc, but I wanted a more detailed explanation than this if that's possible?

Thanks!

Replied on Mon, 02/11/2019 - 09:00
Picture of user. jpritikin Joined: 05/23/2012

A weight of 1.0 counts as a single observation. A weight of 2.0 is like you entered the same data twice as independent rows. A weight of 1.5 is halfway between these two cases. Make sense?
Log in or register to post comments
Replied on Mon, 02/11/2019 - 09:09
No user picture. ZR Joined: 07/26/2017

Thanks!

Ok so it's just like normal weighting. I wasn't sure if it did something more complicated under the hood, but that's great. Thank you for clarifying!

Log in or register to post comments
Replied on Mon, 02/11/2019 - 09:46
Picture of user. jpritikin Joined: 05/23/2012

Well, it is a bit more complicated because fractional weights only work correctly for the maximum likelihood fit function. It doesn't work quite right with weighted least squares. On the other hand, row frequency (an integer) works in more cases.
Log in or register to post comments
  Login

News