Comments on: Ratings readies for Live & Time Shift

By: Tex

Tex — Wed, 09 Sep 2009 01:00:59 +0000

@Neon Kitten: Funny, one of the times I was selected was when I was single and living in an inner-city apartment. I bought my Toppy the afternoon they collected the diary…

Experiment / survey design is a bit of a non-obvious science on the face of it. Ask most people and the answer would be “random sampling, of course!”, but it’s not that simple if you have to account for things like selection bias, etc.

For example, contacting (say) 10,000 people by phone between 9am and 7pm and randomly choosing 3035 of those, would probably lead to a sample weighted towards stay-at-home mums and unemployed people and against professionals or singles. It’s arguably hard to get away from all forms of bias; about the only thing you can do is accept that you’ll always have some and codify it in some way i.e. use demographics, and set sub-sample sizes based on proportion relative to the whole population.

That said, it’s a bit silly to see 7 & 9 fighting over 0.1% of market share when the confidence interval of results is +- 1.75%

By: Jerome

Jerome — Tue, 08 Sep 2009 08:31:33 +0000

the problem that OzTam faces is that unless there are 22million boxes, people will always say it is not enough. so all we can do is look at relative numbers to overseas systems and yes in that regard they do quite well.

By: Neon Kitten

Neon Kitten — Tue, 08 Sep 2009 08:17:03 +0000

Well, I’ve been rejected simply for living alone in a CBD apartment. Didn’t even get to the time-shifting.

I don’t count when it comes to ratings. That’s why I time-shift, ad-skip or see pretty much everything elsewhere these days, it makes no difference

By: slydoggie

slydoggie — Tue, 08 Sep 2009 07:37:34 +0000

I think this is one invasion of privacy that most people would be happy to accept as no-one likes other people speaking on their behalf (I realise that statement is an oxymoron!)

Unfortunately though I think there’d be those out there who would also think it’s pi** funny to tune into Mass for you at home or something random like that.

By: JM

JM — Tue, 08 Sep 2009 07:02:09 +0000

@Jed – Digital boxes are receivers….. not transmitters. So that wouldnt work

By: David Knox

David Knox — Tue, 08 Sep 2009 07:01:01 +0000

Correct. It’s a cross section.

By: JM

JM — Tue, 08 Sep 2009 07:00:14 +0000

So if you go by Tex’s figures, then 1 surveyed household represents 2502 households (altho the article does say Metropolitan – so it may not be 100% accurate)
But how many “people” does one household equal? or are they all different depending on who lives at the house?

By: Jed

Jed — Tue, 08 Sep 2009 05:50:20 +0000

I’m no tech head but with digital boxes being in a large number of homes already and soon in all… wouldn’t it be easy to get an accurate rating figure by going on the digital boxes??

By: Russell

Russell — Tue, 08 Sep 2009 01:17:58 +0000

The US has 5000 Neilson boxes for a population of 310 million.
Or one for every 62,000 people.
Australia has one for every 7000 people. So in our way our system is almost 10 times more accurate.

By: Tex

Tex — Tue, 08 Sep 2009 00:37:05 +0000

Population of 7596000 households (2006 ABS figure), sample size of 3035 households, gives a confidence of 95% +- ~1.75%.

Even if you calculate it on raw population figures, for a population of 22 million at the same confidences you end up with a sample size of 3136 – note, not “households”, but “people”.

Can’t comment on the “hand-picked for suitability” allegation, except to say that apart from seeing it bandied around all the time, I didn’t notice any evidence of it when I was chosen a couple of times years ago. More recently I’ve been rejected because I watch almost 100% of TV time-shifted, but that’s the problem with their measurement methodology which they’re trying to fix, not with their survey design or sample choice.

That said, I don’t particularly believe them either, but that’s because I’m a distrustful bugger

By: Andrew B

Andrew B — Tue, 08 Sep 2009 00:36:04 +0000

So when you see something like: The Nanny GO! 62,000 26,000 10,000 13,000 11,000 1,000 – based on 3035 people meters, a figure of 1000 would mean that one person in Perth with a people meter flicked past the channel and maybe stopped for a few minutes.

The accuracy of the ratings must be very bad when you get below 100,000 viewers.

Although I have read that OzTam regularly check to make sure the sample size is adequate to provide a reasonably accurate picture of what we are watching to a margin of error of about 1%. (not quoting, this all from memory).

By: James

James — Tue, 08 Sep 2009 00:23:12 +0000

I believe Australia has the largest sample percentage v population out of any other country.
The U.S has 25,000 boxes for a population of 300 million.

By: GuanoLad

GuanoLad — Tue, 08 Sep 2009 00:04:35 +0000

I find it hard to believe that only 10-15% of the population watch TV at any given time of day. That doesn’t sound right to me.

By: hoople

hoople — Mon, 07 Sep 2009 22:29:53 +0000

hmm neon kitten – have you seen the sample sizes of other countries?
Australia actually has a pretty robust sample compared to other nations

as for the sample representing australia – it is actually a “mini” australia in terms of age/gender/ region

By: Craig

Craig — Mon, 07 Sep 2009 20:48:03 +0000

welcome to 2007….

By: Neon Kitten

Neon Kitten — Mon, 07 Sep 2009 20:39:57 +0000

Still just as invalid as always, and just as designed to get the results OzTam’s owners – the networks – want.

3035 homes is not an adequate sample for a population of over 20 million – especially given the proliferation of new channels and technologies.

And of course, the sample is hand-picked for suitability, rendering the results even more useless. It’s like taking an exit poll at an election but dismissing any respondents who say they vote Labour.