The FastDay Forum

[Moogie]

Sorted now Lil Perhaps I should look at enabling decimals in the pounds field. Shouldn't be too much trouble.

[carorees]

Hi Moogie, a coupe of people are saying they've input data but is not showing up in their profile...think it was ujs and Claire007 (posted in FAQ section)

Hi Moogie,
great job, thanks for all the work! A suggestion - how about adding the possibility to register if and on which days you actually fasted? Some weeks it might not be possible to do both days, or motivation may be reduced. It would be interesting (and also motivating) to see the correlation between actual fasts and weight and waist.

[Moogie]

Thanks Caroline, I'll have a look and see if I can find what the problem is for those members.

Coffecat: Thanks for your suggestion! I'll have to ponder on how I might be able to add that and show it somehow with the graphs. Can't promise anything but will certainly consider it when time allows!

Moogie, something seems odd with the ADF figures which at present show 3 members with an average weight loss of -4 kg, that's an alarming (and implausible) weight gain of 4kg. I guess someone put some info in wrongly, but it does raise the question of whether data needs to be 'cleansed' somehow before it becomes part of the overall stats i.e. an individual's data that is clearly wrong is excluded.

Wouldn't that become an admin nightmare timewise?

If it was done manually yes, but I was thinking of some automated process. This couldn't guarantee to pick up every mistake but it could pick up egregious ones (that might otherwise skew the overall data). A (better?) alternative would be to look at medians rather than means, but this would also involve some coding.

[carorees]

I think you're right dominic, it would be good to have some way of questioning outliers, but how?

Now I think about it, using medians would be a better way.

Nerd alert! Risk of telling some people what they already know and boring others with stuff they never wanted to know, apologies in advance...

If we are only interested in averages then the best way might be to look at the median results instead of the mean. To work out the median, take all the results for the given statistic (e.g. weight for all those on 5:2), and instead of adding them up and dividing by the number of participants (the mean), you put them in line from lowest and highest and use the result that is half-way down the list (the median). Of course the computer does this, not Moogie nor a busy forum moderator!

The big advantage of a median average instead of a mean average is that 'outliers' (including mistakes or someone deliberately trying to skew the data) have negligible effect because in practice the middle results are always clustered around a very similar number (a 'normal distribution'). So maybe no need to cleanse the data beforehand, which also avoids any implication (however false) that the data has been 'improved' to give 'better' results.

I am not sure whether the code that Moogie is using offers the possibility to use median instead of mean, if it does then it would be interesting to try it?

[carorees]

You are right, of course, dominic to suggest we use medians. I'm sure that it could be coded easily as it is such a standard statistical method.

[Moogie]

I'll try and have a look at this when I have the time. At the moment it loops through the data, adds up the losses (start weight - current weight) and divides by the cumulative number of weeks (calculated from users start dates to last weigh in date). Not quite sure how I'd do a median with that, as I said I'm no mathematician!

The simpler option would be for me to not add on any negative losses, since that's not really a loss anyway, and to check w:h ratio isn't below some silly number.

[Moogie]

Just had a quick look, the problem there was that someone had tried to edit their starting stats through their weigh in record, and that was throwing things off. I've now disabled the ability to edit the first stat, though I will need to go and check through the data to ensure that no-one else has done this too. Don't know why I didn't make this un-editable in the first place!

Hi Moogie,
oops-I hope that wasn't me....but I don't think it could have been as I have only just tried to do it in the last five minutes....? Anyway, I tried to edit my start stats to add my waist measurement which I didn't initially add- is there away to do this ? Thank you !

[Moogie]

Nope, wasn't you my dear - but even if it was, I wouldn't 'name and shame' I take the blame, I should have coded around that happening in the first place.

Anyway, please PM me with the measurement you'd like added in and I'll do that for you asap.

[carorees]

I had another thought...it might be useful to be able to add to the filter settings the ability to filter by time following the diet. As time goes on, more people will be entering the weight maintenance phase which will result in the average weight lost per week to go down. If we could filter by the same time durations as in the pie chart, it would give a better picture of what to expect?