Re: LME- log-normal distribution of parameters

[Spencer Graves <spencer.graves@pdf.com>]

     I agree you need to worry about conservation of mass, etc.  
However, it also doesn't make sense to expect the same margin of error 
on dollar terms between $1 and $1B.  Mony and physical measurements are 
often lognormally distributed, and it often makes more sense to think of 
relative error than absolute error in modeling. 

     In electricity, power = current * voltage = 
(current^2)*resistance.  If you don't know this but have crude 
measurements of power, current and resistance, a linear regression on 
the decibel scale will estimate the powers AND a reconciliation of the 
units. 

     spencer graves

David Parkhurst wrote:

> You also want to be very careful, when transforming data to logs, that 
> the logarithms make sense scientifically.  In situations involving 
> amounts of mass or energy, they often don't make sense.  There is a 
> law of conservation of mass, but no such law for logarithms of mass.  
> Next time you reconcile your monthly checkbook statement, try 
> converting all amounts to logs first---things just won't add up.  So 
> think this through for your own data.
>  
> Dave Parkhurst
>     ----- Original Message -----
>     From: Pravin <mailto:jadhavpr@vcu.edu>
>     To: david.thompson@mnr.gov.on.ca
>     <mailto:david.thompson@mnr.gov.on.ca> ; s-news@wubios.wustl.edu
>     <mailto:s-news@wubios.wustl.edu>
>     Sent: Friday, March 26, 2004 11:36 AM
>     Subject: Re: [S] LME- log-normal distribution of parameters
>
>     log(predictor+0.0001) would be the simplest and
>     intuitive solution. But for some reason I need to do better than
>     that.
>      
>     Looks like glme() in the library(correlatedData) offers such
>     flexibility but my experience in this field is almost equal to
>     none. So looks like some background work is needed before
>     implementation.
>      
>     Thanks for all the help. 
>      
>
>     Thanks,
>
>     Pravin
>
>     Pravin Jadhav
>
>         -----Original Message-----
>         From: david.thompson@mnr.gov.on.ca
>         <mailto:david.thompson@mnr.gov.on.ca>
>         [mailto:david.thompson@mnr.gov.on.ca]
>         Sent: Friday, March 26, 2004 11:21 AM
>         To: jadhavpr@vcu.edu; s-news@wubios.wustl.edu
>         Cc: slarsen@insightful.com
>         Subject: RE: LME- log-normal distribution of parameters
>
>         Does log(predictor+1) seem a reasonable alternative?
>          
>         DaveT.
>         **********************************************************
>         David J. Thompson
>         Silviculture Data Analyst
>         Ontario Forest Research Institute
>         Ontario Ministry of Natural Resources
>         1235 Queen Street East
>         Sault Ste. Marie, Ontario, P6A 2E5
>
>         (705) 946-7433
>         (705) 946-2030 Fax
>         david.thompson@mnr.gov.on.ca
>         **********************************************************
>          
>
>             -----Original Message-----
>             From: Pravin [mailto:jadhavpr@vcu.edu]
>             Sent: March 25, 2004 6:42 PM
>             To: s-news@wubios.wustl.edu
>             Cc: slarsen@insightful.com
>             Subject: LME- log-normal distribution of parameters
>
>             Hello,
>
>             >lme(response ~ time, data=data.g, random = ~ 1+time|ID)
>             ##"time" is used a predictor for the "response" in the
>             data frame "data.g"
>             ##Random regression intercepts and slope on "time"
>             ##Random effects vary over  individual ID
>
>             The above model assumes that the intercept and the slope
>             are normally distributed. How can I specify log-normal
>             distribution for these parameters. But the residual error
>             can be normally distributed-- that is fine. One suggestion
>             was to use log transformation. But I cannot use log
>             transformation in the fixed effects model
>             (log(response)~log(time)) because there are a few 0's in
>             the predictor column.
>
>             Thank you,
>
>             Pravin
>
>             Pravin Jadhav