Monday, 24 February 2014

stats aren't facts

So, with today's banter with Neil McRae on twitter I did finally get :-

"stats aren't facts though - I wish they where sometimes! :)"

Now I wonder. If they are not facts, what the hell are they?

Of course stats are facts. We have facts like "At time X we sent 100 LCP echos and received 97 replies". That is a stat, and a fact. An absolute irrefutable fact. These "stats" are the raw data from the tests done. The facts we start with to try and make sense of the stats.

Forecasts are not facts. Some of the interpretation from stats are not facts.

But really, Neil, stats are facts, hard solid facts, and if you don't know that then, well, we have a problem

I am also accused of bullying BT on the whole congestion issue. Fair point, but if Neil would like to tell me the correct (and effective) way to get BT to take such issues seriously, I'll be more than happy to do that instead.

Update: Neil has hidden and/or deleted his tweets. Ones where he said BT customer services would agree with anything to keep me happy, where he said it was appropriate to call me stupid, where he says stats aren't facts. Oh well.

8 comments:

  1. > I am also accused of bullying BT on the whole congestion issue.

    Funny - if you were just a residential customer, they'd give you a standard BS response and laugh in your face. Soon as you got a voice, you're a bully.

    Sucky sucky.

    ReplyDelete
  2. Probably the real correct answer is to start migrating user circuits over to TalkTalk backhaul wherever possible - which, as long as you don't fill up their core faster than they can keep up with, would mean reducing the congestion on congested paths, reducing BT's revenue and giving a better service to end users too: a win for everyone except the "our packet transport failing to transport packets is not a fault in our book" camp.

    Dropping LCP echoes on an otherwise-idle line means the line is failing to achieve throughput of 80 bits per second - yes, bits, not megabits or even kilobits. Would FTTC sell well if marketed as "may fail to achieve 0.1% of the speed of ISDN"?

    ReplyDelete
  3. No wonder BT's network is in such a mess if the man at the top come out with such rubbish

    ReplyDelete
    Replies
    1. indeed, I am 100 miles from london.
      An example of NT's wonderful wholesale network.

      latency to bbc.co.uk on fast path dsl services

      BTWholesale 17-22ms (it varies as seems to have alternate routes)
      Easynet(sky) 9ms
      Talktalk 14ms
      C&W 16ms

      and on cable via virgin media 13ms although due to congestion thats just the base latency actual latency jumps all over the place.

      Reason for high BT wholesale is I believe all my traffic is diverted north first via sheffield.

      Delete
  4. @keithb - The BT network is not 'in such a mess' - obviously it's imperfect, but that shouldn't be surprising.

    Neil McRae is a competent guy running an enormous network and who has a reputation for robust public communication. RevK is much the same, except that his network is several thousand times smaller.

    ReplyDelete
    Replies
    1. Indeed, but I would not simply say someone is being stupid. I'd look at the evidence.

      Delete
  5. Nicholas has left a new comment on your post "stats aren't facts":

    statistics (stəˈtɪstɪks)
    noun
    1.
    the practice or science of collecting and analysing numerical data in large quantities, especially for the purpose of inferring proportions in a whole from those in a representative sample.

    So, assuming that the methodology by which the statistics are collected is sound, the statistics are, indisputably, absoute facts.

    inference (ˈɪnf(ə)r(ə)ns)
    noun
    1.
    a conclusion reached on the basis of evidence and reasoning.

    So, it's the inferences drawn from the statistics which may not be factual. Giving Neil the benefit of the doubt, perhaps what he meant was "your inferences are not facts"... However, one would expect that anybody refuting an inference would be able to bring additional information to the table to demonstrate the inference is false.

    ReplyDelete
    Replies
    1. I have been following this thread with interest. I do not think Neil would last a second working for CERN (the LHC) as his belief flies right in the face of how they (scientists) obtain stats to infer facts about our physical reality. I am, of course, referring to the hunt for the Higgs Boson, where their stats now infer, with a big fat 5 sigma confidence, that the Higgs is real, and is fact.

      So stats are facts, provided they are interpreted correctly and enough data is collected to have a high sigma confidence in them.

      If BT have employees like this working for them, there is no hope for us. :(

      Delete