[R-sig-ME] Mixed model with counts data

[David Duffy]

On Thu, 30 Jun 2011, Chun (Jimmie) Ye wrote:

> Hi all,
>
> I have a rather peculiar dataset that I'm not sure how to model
> properly. This is data from an instrument that measures the size of
> particles but instead of giving a continuous value, it generates a
> "histogram" of the counts for a particular bin size. So the data looks
> like this:
>
> 	Condition 1				Condition 2
> Dimension	Rep1.A	Rep1.B	Rep2.A	Rep2.B	Rep1.A	Rep1.B	Rep2.A	Rep2.B
> 2	5	6	7	8	40	35	33	31
> 2.1	7	8	4	5	30	30	31	29
> 2.2	10	11	10	12	20	18	21	20
> 2.3	50	45	44	39	5	8	7	7
> 2.4	80	90	75	77	8	10	3	5
> 3.5	30	22	31	35	10	5	7	9
> ?
> 50	0	0	0	0	0	0	1	0

> There are two biological replicates and two technical replicates for
> each of the conditions of interest. If these were continuous, I would
> just fit a nested mixed model to the data.

> [...] a KS test but I'm not sure how to incorporate the replicate
> information (except for averaging). Finally, I thought perhaps I could
> estimate directly from the histogram the variance components from the
> data.

So each single biological sample gives rise to a distribution of 
particle sizes?  Are you interested in central tendency, peak height (ie 
number of particles), shape of distribution?  I would have thought 
treating it as a coarsened continuous measure would make best use of the 
the data (like for coarsened Gaussian, AS319; Wolynetz Appl Stat 1979 
28:195-206).

Just 2c, David Duffy.


-- 
| David Duffy (MBBS PhD)                                         ,-_|\
| email: davidD at qimr.edu.au  ph: INT+61+7+3362-0217 fax: -0101  /     *
| Epidemiology Unit, Queensland Institute of Medical Research   \_,-._/
| 300 Herston Rd, Brisbane, Queensland 4029, Australia  GPG 4D0B994A v