Evolution

[Karl Larsen]

Leif Gregory wrote:
> On Mon, 28 Jul 2008 17:53:51 -0600
> Karl Larsen <k5di at zianet.com> wrote:
>
>   
>>     OK your such a lier and think telling lies make them right. So 
>> please do exactly what I do. Look at the results and give us all the 
>> results. Then try to make it look like your stuff.
>>     
>
> I can't quite make out if you're asking me to give you statistics on
> where you were wrong or not.. But if it'd make you happy send me an
> e-mail off-list and I'll do it for you. Not sure it's going to make
> your life any better for it.
>
> So let me put this another way. If you want people to take your "work"
> seriously, then you need to do it seriously. You need to come up with a
> methodology that is going to give you results (whether they match your
> expectations or not, the whole, lies, damned lies, and statistics
> stuff) in a manner that is as accurate as possible and is repeatable by
> others.
>
> Here, I'll even get you started.
>
> 1. You need a way of parsing through messages to extract the MUA
>
> 2. You need to dedupe those results so that you don't count say, me
> more than once for using Claws Mail. This is a problem in your current
> methodology.
>
> 3. There are thousands upon thousands of Ubuntu users out there, and
> there is a reasonable expectation that the majority of them use e-mail.
> Your problem is to find a data source that captures a very good cross
> section of those users. This is another flaw in your methodology.
> Here's why.
>
> a. Some users on this list are much more vocal than others, therefore
> the chance that your sample (which you've decided on 100 messages)
> contains a greater weight of the vocal users is more likely. See number
> 2. Therefore, your methodology must include a way to get a sample of
> 100 messages with no dupes. That means you're most likely going to have
> to parse more than 100 messages to get your sample. This brings up
> point 3b.
>
> b. Your methodology is flawed in that you're sampling only messages, not
> users. Some people here have already mentioned that they use different
> clients based on where they're at. i.e. Apple Mail at home, Thunderbird
> at work etc. You're going to need to account for single users who use
> more than one client. If you don't it will skew your results. You're
> trying to show what the most popular MUA, and those folks will throw
> off your results that could potentially degrade your case.
>
> 4. Is this list the right place to take your sample. Probably not. I
> don't believe it's truly representative of the Ubuntu population at
> large. If I want to sample the homosexual population as a percentage
> of the overall population, my results are going to be completely
> different if I take my sample in Albuquerque vs Santa Fe (where the gay
> and lesbian community is much larger). You need to find a place you can
> get your sample that reflect the Ubuntu community at large.
>
> So why is a list not representative? I ran five discussion lists like
> this one from 1997 till early 2007. One thing we discovered is that new
> users typically flocked to forums first rather than e-mail lists.
>
> We even watched some long-time members of the lists go to forums
> because it was easier to find their answer due to the persistence of
> the threads in a searchable manner.
>
> So, you have to ask yourself, which population do you want to sample
> from.
>
> Well, there's the basics Karl, if you can't understand where you went
> wrong from the above, then do what you're doing now but don't come back
> whining that nobody takes you seriously and Evolution doesn't get
> dropped as a default install.
>
> Now if you really want to do it right, make a strong case with
> verifiable facts, then read up on statistical analysis. Google will
> nets you tons of great info. I'm not going to be able to help you in
> that arena, way too much math for me.
>
> Here's one to get you started with regards to an acceptable sample
> size. This will at least tell you if your 100 messages is sufficient.
>
> http://stattrek.com/Lesson6/SampleSize.aspx?Tutorial=Stat
>
>
>
>   
    Your building a case for errors to occur due to the sample size and 
the source of the data. I have never said it was perfect. Yes there are 
all the problems you point out.

    But are you saying that in the general Ubuntu population use of 
Evolution is greater than the use of Thunderbird? That makes the users 
of this list weird. And it is not supported by any statistics I have seen.

    The other method that gets this result seems to use how often the 
software is downloaded. This gives Evolution a big edge because it is 
not downloaded and the number of users are those who didn't download 
Thunderbird. This has real error built in.

    This was started because there are a lot of people wanting to remove 
Evolution and have problems after doing so. It is a real problem.

Karl


-- 

	Karl F. Larsen, AKA K5DI
	Linux User
	#450462   http://counter.li.org.
   PGP 4208 4D6E 595F 22B9 FF1C  ECB6 4A3C 2C54 FE23 53A7