Mrs Barista, well done this must be the most mind numbingly boring thread every

Grubbs

Thanks bud - up against some fierce competition as well


Grubbs is my new favourite.

Wahey! Now we're getting somewhere. I have to admit, that's a bit better than excluding Quins because in your opinion they distort the figures.



Unfortunately, Quins are undoubtedly part of the population of SL clubs, this is self evident and as such the Grubbs test is neither needed nor justified. You're not testing for an erroneous result, the financial results of a Premier League football club haven't slipped in by mistake. Your data set is the financial results of SL clubs.

"One possibility is that the outlier was due to chance. In this case, you should keep the value in your analyses. The value came from the same population as the other values, so should be included."

[urlhttp://www.graphpad.com/quickcalcs/GrubbsExplain.cfm[/url







Close but no cigar. Still, fortunately for Mr B you'll now be busy all weekend trying to justify your use of Grubbs so at least he can watch what he likes on telly. You're welcome Mr B.

I would be happy to let that go, if my better half liked rugby. But she doesn't, and she has to have the last word.

Unfortunately for you, Grubbs test showed that the value for Harlequins was a statistically significant distance from the rest of the population so I will exclude it, thanks. In fact you might say that it backs up non-expansionists like Roofs who who have long held the view that Harlequins inclusion in this population is erroneous anyway and Grubb has provided some empirical evidence to support this opinion.

That is just total nonsense and proves you have no idea what the technique is for or what the results show.

In any event you are using the test incorrectly. Perhaps if this were the first set of SL clubs accounts you could argue the Quins figure is exceptional and might be disregarded on that basis. But it's not even exceptional for Quins, it's a fairly normal result.

What you have done is to collect a set of data, used that to make some kind of point, seen that including Quins weakens your point and searched for reasons to validate the exclusion because in your opinion it doesn't belong. This resulted in the patented Mrs Barista prove what you want from any data set methodology by picking an arbitrary number that includes the data you want but excludes those you don't.

You then tried a different approach arguing you were merely pointing out that 4 clubs were profitable compared to one before and claimed this was the real point you were making. You don't need any sort of made up statistical analysis to prove that, we can see from the list of figures you supplied.

Then you tried to argue Quins could be excluded because their loss was 2 x greater than the next biggest loss making club. You ignore the fact that Leeds profit is 4.5 times that of the next biggest profit making club.

Finally you sought salvation in Grubbs (quite apposite

) but use the technique selectively and not in the way it should correctly be used because you think this justifies your reasoning. It doesn't.

Quins are part of the population of SL clubs no amount of erroneous use of statistical techniques will change that. Their result falls within +- 3 standard deviations of the mean, which is where we would expect 99.7% of the data. Their result is normal for the data set studied.

There is no basis to exclude them, to do so would be to introduce an obvious (to anyone but you apparently) bias into the findings.

Does anyone have the number for the Samaritans?

At the Wigan End of Season dinner last Tuesday Ian Lenagan said the club was now in profit. No doubt year starting Oct 2010/Sept 2011.
Considering where he has brought the club from financially he's done an exceptional job. With Silverware in the trophy cabinet too.

To suggest that I suddenly brought in the "4 clubs in profit" point late on is utter nonsense but I understand your desperation having asked for an accepted methodology for outlier identification to be applied and it proving one of my initial points. My opening post started with this: "Seems like 2009 was a marginally better year financially for some clubs. In 2008, the only club (I think) to make a profit was Hull FC. The league table of profit/loss for 2009 is as follows (assuming I've got the right entities) and shows 4 clubs in profit and a further 3 with a loss < £100k". So far from this being some hastily thrown in point, it was actually my opening comment.

You challenged me excluding Harlequins on the basis of my opinion and requested an accepted methodology for identifying outliers. The most recognised technique for this is Grubbs, which, when the data set was run through the calculation, showed Harlequins were indeed a statistical outlier. You didn't like this and suggested an alternative formula, not one specifically designed identify a single outlier in a data range, I should add, which was precisely what you asked for (that test being Grubbs, obviously). Getting a bit huffy, having asked for a methodology and been provided with one that validated my initial view, you suddenly abandon your need for any statistical model because now I'm told there can't possibly be one. You should make your mind up pet. You resort to saying the point about 4 clubs being in profit was some sort of desperate last ditch attempt by me to salvage the argument. I've now corrected you with a reference to my opening post. Grubbs rules, I win, Harlequins are excluded.

Good news from Wigan that they have returned to profit in 2010, so more evidence to back up my first point that maybe things are looking up a bit for SL clubs. Plus if Widnes have made the hefty profit reported here they would presumably replace a weaker club from a financial perspective in the next round of franchises which would be another step on for SL financially. Good news, I'm sure you'll agree.

I'm just firing off a letter to Wisden instructing them to recalculate Graham Gooch's Test average excluding his brazenly un-average 333 versus the Indians in 1990.

Send me a list of his Test scores and I'll run them through Grubb's model to see whether 333 is a single value outlier in that dataset.

saddo. can back up your argument but still very very sad, i'm a rugby fan and like to watch a game of rugby not use a freakin abacus

She is like this every frigging day on the HKR site.
To say it gets boring is a complete understatment.

Quote ="Mrs Barista"Send me a list of his Test scores and I'll run them through Grubb's model to see whether 333 is a single value outlier in that dataset. '" stats.cricinfo.com/ci/engine/pla ... ew=innings

Quote ="Mrs Barista"Send me a list of his Test scores and I'll run them through Grubb's model to see whether 333 is a single value outlier in that dataset. '" stats.cricinfo.com/ci/engine/pla ... ew=innings

Cheers bud. I'm a messageboard fan and like to post without personal insults. In my defence it was your chum Barnacle Bill who started this averaging/outlier exclusion extravaganza, which unfortunately has derailed what was intended to be a positive thread about improving financial performance in some SL clubs. I'm not sure why this is TBH.

Exactly right.

Within these sports club models profit is almost irrelevant. What is most relevant is the operating cashflow, as it is that which dictates how much the club benefactors need to inject.

If Mrs Barista has all of the club balance sheets (opening and closing) then perhaps she can construct a simple FRS1 cashflow statement and tell us what their operating cashflow is (as opposed to net cashflow) ? That is a much better indicator of how clubs are performing than arbitrary P&L reserve movements.

Profit is vanity, cash is sanity.

Eh? Don't you need to have disclosure of cash equivalents (ie overdrafts) to do that properly?

I challenged excluding Quins "in your opinion", you then suggested a "standard deviation diagram". I didn't need to invent one of those because by calculation it is clear that Quins fall within +- 3 standard deviations of the mean.

You decided to persist with the nonsense of that approach and made a graph with an arbitrary cut off £1,000,000 above and below the median as if that proved anything at all.

At this point you included this strange statement "It's almost as if you're trying to downgrade the financial performance of the rest of the competition so the results of your club are compared to an average distorted by a loss twice as great as the next worst club. Assuming the accounts are filed at some point." Which I think is telling in terms of the real purpose of this pseudo scientific excercise.

Even in this statement you cling onto the notion that because Quins loss is twice that of the next biggest loss making club that in some way justifies their exclusion. You still have not explained why that logic should not apply to Leeds whose profit is 4.5 times that of the next biggest profit making club.



I haven't abandoned a request for valid statistical technique which reasonably excludes Quins at all, I'm still waiting for one.



The Grubbs test cannot exclude Quins because Quins are a part of the data set "SL Clubs profit/loss 2009". Their results are normal for that data set. They fall within +- 3 standard deviations of the mean, which is where we would expect the vast majority of results to lie. They do.

You will continue to grasp at this straw though because having made an absurd statement "In my opinion Quins are not representative" you then sought some kind of technique (after making up a couple of your own) which, if used inappropriately, appears to support your exclusion of Quins. Grubbs doesn't, Quins are a SL club and their financial results contribute to the data set "financial results of SL clubs 2009"

If, as has been mentioned, Crusaders post a £1,000,000 loss what happens to Quins then? Do they get put back in your calculation because they've suddenly become a SL club again?

What if Saints and Hudds had not reported yet or you didn't like their results either so decided to exclude them "in your opinion" as well as Quins. Do you then also exclude Leeds because they become an outlier? And then what is the value of your conclusion about the data set?

The reason for this preparatory "study" concluding in a made up "average" is to pave the way for a further dig at Rovers when their "less than expected" losses are published.

If people think that unlikely then they only need to see a few of your posts on the Rovers forum.



You may follow Hull FC but your passion is sniping at Rovers.

Mrs B is certainly an-a-Count-ant.

What, the average that was mentioned once until you started this stats-fest you mean?

Neil Hudgell reported that Rovers lost £450k last November, so why would actually filing those accounts make any difference to what I've already posted, other than taking the average down a bit? Unless the "improvement" he referenced is in the region of +£250k it won't change Rovers relative position to the other clubs. It won't alter the fact that 4 clubs have turned in a profit this year vs 1 last year and a further 3 are around(ish) breakeven, of which one is Wigan who have confirmed a return to profitability this year. It also won't change the fact that Harlequins are a statistically significant outlier from the rest of the clubs who have filed. The point you make on Saints/Hudds/Leeds being excluded on a whim is a great one, because if you run Grubbs against data for everyone excluding Harlequins, there are no significant outliers in this population that may be excluded.

I've never talked about club x's profit / loss being greater or less than a certain multiple of the next nearest, that was you, cockle. A shame then that this was shot down by the key statistical test for determining outliers.

Using a separate argument, just in case you now say I'm trying to muddy the waters, I do commercial finance for a living. I do it pretty well and get paid pretty well for doing it. One of the key things I look at is relative performance, whether by geography, channel, brand, category, you name it, and I define control groups on all these dimensions on a daily basis. I can tell you now that from a performance review perspective I'd be discredited if I reported a benchmark that included an outlier of this magnitude. I could report both, but there would be a demand to strip one out also to see a more representative average performance.

You seem rather unwilling to discuss the key point here which is that on the face of it, results generally speaking are a bit better. I think this is a good thing, although as Derwent correctly points out, a bit of a crude measure, but nonetheless one that has been used in the past. Do you think it's encouraging? Even if include Harlequins, add in £450k loss for Rovers (this being worst case given the improvement still to be added in) and £1m for Crusaders, the average loss is -£370k, so an improvement. But the real improvement surely is from a position in 2009 of only one club in profit and Hudgell's Sun article saying "everyone is losing £500k a year", it seems perhaps we are now in a better place than that. Or am I being too naive?

Who said it would?



And if you exclude Saints and Hudds "in your opinion" Leeds becomes an outlier. Does that mean they're no longer a SL club suitable for inclusion in the data set 'SL clubs finacial results 2009'? What this does is reveal the nonsense of your method. As does including a loss (not confirmed) for Crusaders of £1,000,000 because then Quins are no longer an outlier using Grubbs technique.

But I stress, the use of Grubbs is erroneous, again you choose to ignore this because it highlights your lack of statistical rigour. Grubbs would be used to check if a result belongs in a data set i.e it is not an erroneous result from a different data set that has crept in by mistake. Unless you are arguing that it is a fundamental mistake to include Quins in the financial results of all SL clubs, which might be the case if say the results of the Union club had been used or some other error had been made. But that is not the case.











Well done you.





You wouldn't, because you've only just learned about Grubbs. If you used this test (wrongly, presumably since you are here

) in your everyday work, you wouldn't need to exclude Quins "in your opinion" in the first place, or make up some arbitrary cut off point around the median, when I questioned your method.





If you include Crusaders supposed loss of £1,000,000 you HAVE to include Quins by your own methodology. This is the absurdity of your position. Quins aren't a SL club but if Crusaders are included then Quins become a SL club again.



Your hanging off Neil Hudgell's every word is not really a revelation.

Not at all.

If you have the balance sheet you can work out the OCF quite easily from the movements in working capital and the annual EBITDA. So long as you have the opening and closing balances for debtors, creditors and stock then you can work out what the cashflow was from normal operating activities though in some cases you'd have to have an educated guess at EBITDA due to lack of a detailed P&L.

You could exclude Hudds and Saints because it makes about as much sense as excluding Quins.



I asked for a better reason to exclude Quins than "in your opinion" you invented a couple including "Look at the chart. Look at the banding of 10 clubs within the two lines showing £1m and -£1m, also +/- £1m around the median." Which proves absolutely nothing. "Look at the chart" FFS!



Then you stumbled upon Grubbs and decided it solved your problem, you don't know what it's for or how to use it correctly do you? Oh hang on I forgot!









Are you seriously suggesting that Quins are not a part of the data set SL clubs? Are you saying that they do not fall within +- 3 standard deviations of the Mean? Exactly where we would expect the vast majority of clubs results to fall?

If you are, I hope your boss doesn't know your user name on here.







On average every club was losing £500K that is what an average tells us, the average result for the whole data set. There is likely no single result that exactly matches the average, some will be higher and some lower. Some may even be a lot lower but that does not mean they should not be included regardless of whose opinion it is.



On a like for like basis (i.e. not excluding any given club for spurious and statistically false reasons) if the average is less of a loss this year compared to last, then of course (bearing in mind the many ways financial results can be massaged) that is good news.

What we can't do is compare an average using a spurious and statistically incompetent method this year against an average calculated in a different way for last year.

Quite easy with a full set of accounts. Much harder with abbreviated accounts, especially where you have non-recurring transactions or financing activities that are not disclosed.