Field Theory by Hadyn Green

Read Post

Field Theory: On Averages

48 Responses

First ←Older Page 1 2 Newer→ Last

  • Graeme Edgeler,

    Those really shouldn't be line graphs :-)

    Wellington, New Zealand • Since Nov 2006 • 3215 posts Report

  • Hadyn Green,

    Those really shouldn't be line graphs :-)

    Yes they should. While they are discrete data points. I am attempting to show change over time. This is also why missed games are not represented by zero. (Argued this with Amy yesterday)

    Wellington • Since Nov 2006 • 2090 posts Report

  • Hadyn Green,

    Though I will admit there should be something for the Hong Kong and Samoa games.

    Wellington • Since Nov 2006 • 2090 posts Report

  • Naly D,

    Hadyn, I'm lazy.
    Do you know who was William's partner for those matches? I've noticed he seems to take a different role depending on whether Boric or Thorn is on the field.

    Wellington • Since Sep 2008 • 307 posts Report

  • Hadyn Green,

    Do you know who was William's partner for those matches? I've noticed he seems to take a different role depending on whether Boric or Thorn is on the field.

    That is a very good question. I can look at that.

    Wellington • Since Nov 2006 • 2090 posts Report

  • Graeme Edgeler,

    Yes they should.

    Ooh, this will be fun ... no they shouldn't.

    Change over "time" could just as easily be shown by the heights of a bars on a bar graph, lest we think that on the Wednesday of the week following some game Carter was still doing better than he was on the game on the following weekend...

    Wellington, New Zealand • Since Nov 2006 • 3215 posts Report

  • linger,

    Not so much fun: It's not a question of "should" or "shouldn't", but rather, what interpretive assumptions Hadyn's making about sequential games. A linear graph makes sense if it can be assumed that player ability does "develop" over time; and at least for some aspects of performance or fitness, it might make sense to consider levels between games. On the data available, this assumption seems to work better for some players than for others -- so, the suitability of the analytical frame is still an open question.

    Tokyo • Since Apr 2007 • 1944 posts Report

  • Hadyn Green,

    Change over "time" could just as easily be shown by the heights of a bars on a bar graph

    Hmmmm, interesting ... interesting ... but I would argue: "are you high?"

    Change over time on a bar chart? I get what you mean regarding the time span in between. But I would use a points rather than bars as bars cloud multiple data points. The lines between are to highlight the difference between the points.

    Wellington • Since Nov 2006 • 2090 posts Report

  • Lyndon Hood,

    think it depends whether the graphs represent "number of tackles" or "tackliness".

    My suspicions is that 11 games might be enough to draw a sensible average from but perhaps not a (statistically) meaningful trend.

    Wellington • Since Nov 2006 • 1115 posts Report

  • Graeme Edgeler,

    Change over time on a bar chart?

    It's not really change over time (hence my oblique reference to "time") I had understood it to be change between one match and the next.

    Grouped, coloured bars might be a way to go - a red a blue and a green bar with a big gap between them and the next set of red, blue and green bars...

    Wellington, New Zealand • Since Nov 2006 • 3215 posts Report

  • Lyndon Hood,

    Change over time on a bar chart.

    Admittedly your last one would be rather hard to read done differently but the don't make it right.

    Thing is, lines-in-between traditionally represent continuous change in levels. So 'should' be a line graph where 'should' does not refer to the normal rules of doing graphs.

    Wellington • Since Nov 2006 • 1115 posts Report

  • Hadyn Green,

    Thing is, lines-in-between traditionally represent continuous change in levels.

    Yeah I know, and that's why I would use discrete data points if I was doing it for a research report.

    But wanted to emphaise the zig-zag-iness of the data which I think would be lost in a clustered bar chart.

    btw Lyndon, your link just sent me to a blank page, but it did make me notice that I forgot to put in Dan Carter's chart, will update soon.

    Wellington • Since Nov 2006 • 2090 posts Report

  • Lyndon Hood,

    That would explain why my attempted italic broke. If you want the NIWA NZ historical temperature chart you'll need to replace the <em> with two _'s.

    http://www.niwa.co.nz/__data/assets/image/0006/73455/nztemp7_1853-2007_v3.jpg

    Wellington • Since Nov 2006 • 1115 posts Report

  • Kyle Matthews,

    I would say for it to be a line, there should be uncollected data in between.

    Either, a game every week, but we're only sampling one a month.

    Or, six years of games, and we're showing the average for each year. The line (in theory, it doesn't actually) shows a moving average.

    There is no possible relationship between points, nothing happened between the points where more data could be.

    Since Nov 2006 • 6243 posts Report

  • Kyle Matthews,

    And Haydn, aren't you impressed that we're ignoring all your data and conclusions that you put so much work into, and arguing about what sort of graph you should use?

    Next time you post about the All Blacks, right after they've got the Grand Slam, we'll kidnap the comments into a discussion of the inaccuracies of the meaning of the words 'grand' and 'slam', probably in relation to tennis.

    Since Nov 2006 • 6243 posts Report

  • Kyle Matthews,

    This is also why missed games are not represented by zero.

    I feel that games where players subbed on or off should either not be included, or their numbers should be compensated up in some way.

    After all, if you're on the field for 10 minutes and people were to look at your stats and say "man, you only made three tackles in a whole game of rugby?" you'd be pissy.

    Since Nov 2006 • 6243 posts Report

  • Hadyn Green,

    And Haydn, aren't you impressed that we're ignoring all your data and conclusions that you put so much work into, and arguing about what sort of graph you should use?

    I totally love it! It speaks to a higher cause.

    Though I doubt anyone could argue with my findings: Ali Williams is erratic and Richie McCaw is awesome.

    Wellington • Since Nov 2006 • 2090 posts Report

  • Hadyn Green,

    I feel that games where players subbed on or off should either not be included, or their numbers should be compensated up in some way.

    Good call, I'd be happy to share the data so people can weight it for me.

    Wellington • Since Nov 2006 • 2090 posts Report

  • Hadyn Green,

    Do you know who was William's partner for those matches?

    It was Brad Thorn (as his starting partner) for every game except South Africa in Dunedin and against Scotland. Both times other partner was Boric

    Wellington • Since Nov 2006 • 2090 posts Report

  • Hadyn Green,

    Wellington • Since Nov 2006 • 2090 posts Report

  • Paul Rowe,

    I'm with Haydn, the line chart is way more meaningful than the bar chart.

    Lake Roxburgh, Central Ot… • Since Nov 2006 • 574 posts Report

  • Amy Galvani,

    What is also interesting about the Ali Williams stats is that, while his First 3 stats look pretty wild, there appears to be a bit of predictability in his performance - sharp decreases are usually followed by sharp increase and vice versa. Maybe it's a regression-to-the-mean kinda thing.

    Wellington • Since Oct 2008 • 28 posts Report

  • Jackie Clark,

    I have no comment on the game. I really don't like rugby (although I don't go to the extreme that my mother does, cheering on whoever the All Blacks are playing. Just because. ) However, I couldn't let those beautiful graphs go unappreciated. So. Yes, that's it.

    Mt Eden, Auckland • Since Nov 2006 • 3136 posts Report

  • Naly D,

    It was Brad Thorn (as his starting partner) for every game except South Africa in Dunedin and against Scotland. Both times other partner was Boric

    AH-HA! Mystery solved.
    He went off injured in the English ChCh test and the SA Dunedin match.

    Don't ask me for the reason for the dip in the Eden Park match :/

    Wellington • Since Sep 2008 • 307 posts Report

  • Hadyn Green,

    However, I couldn't let those beautiful graphs go unappreciated.

    Thanks! I love all the pretty-isation options in the new Excel

    Wellington • Since Nov 2006 • 2090 posts Report

First ←Older Page 1 2 Newer→ Last

Post your response…

This topic is closed.