| Topic: | Re:Re:Re:More support than objections | |
| Posted by: | Tom Pike | |
| Date/Time: | 11/08/23 06:22:00 |
| Hi Anette, Thanks for providing some clarification on the six-month figures. In fact the fourth column was cut off in your submission so the information you are now giving was not available from that column header. Your previous column headers included the weeks averaged, confirming you had used 14 non-market Sundays in 11 months of data, (although you had then erroneously averaged by 11) so it was quite a reasonable assumption that you had been consistent here. You previously, but incorrectly, suggested that you made most sales on non-market days, so why not include all non-market days in your latest analysis? And the larger question remains why are you now basing your analysis on just 24 weeks of data rather than the 47 weeks in the original dataset, and on what basis did you chose which weeks to include? Selecting a smaller sample when you have a larger sample available is only going to make it more likely you’ll get unrepresentative results, as well as raise questions as to potential bias in your selection. And of course it won’t capture the seasonal changes that I’m sure have an effect on your sales. For example, on the basis of your original 47-week figures you suggested that your lowest sales when there was a Cheese Market were down to it selling food. That’s a plausible suggestion, but your new figures for just 24 weeks show that your sales were higher on Cheese market Sundays than for the Flower or Antiques market. I don’t think that really proves your initial suggestion was wrong, just that a smaller sample is less reliable. Similarly, your original 47 weeks of sales, when correctly analyses, showed that non-market Sundays gave you the worst sales. A smaller sample might well show the opposite, especially if there was a biased selection, but it doesn’t undermine the analysis of the larger, more reliable, dataset. And of course if sales were higher on your selected non-market Sundays, they would have to be lower, much lower in this case, for the other eight weeks you didn’t select (if you had been consistent and included any fifth Sundays the variation is much less). What makes those eight weeks not chosen any less valid that the six you decided to choose? In the end a larger dataset will always be preferable to a smaller one, giving more reliable results, unless you have some good reasons to exclude some of the data. |