Hacker News Reader

Comment by LaputanMachine | Hacker News Reader

The linked graph may be misleading, because more scented candles are sold in the winter. There is still a correlation if the graph is divided by the number of reviews [1], but it is much weaker.

It would be interesting to see the data before 2020.

[1]: https://twitter.com/nick_beauchamp/status/147355576211076710...

View on HN

Lots of people only leave a review if they experience a problem, so dividing by the number of reviews might not help as much as it seems.

What would be good is to divide it by the number of sales but we don't have that data.

This would have been my assumption too, but the linked charts would suggest that the opposite is true around the holiday season

> It would be interesting to see the data before 2020.

Given the graph you link to starts from January 2020, and it's completely flat until the next winter, where complaints peak in december and gradually decrease until may-june where they reach the pervious baseline, that would be an assertion that 2020 is the odd one out with oddly pungent candles yes?

Unless we think sales or something drop going back before 2020 we are already seeing the key detail, as the peak in the 20/21 and 21/22 boundaries was at the new years but the graph is flat going back to that peak in 19/20.

More than interesting; it would be required to see the data before 2020 to know if this plot is actually meaningful.

Why divide by the number of reviews? They don't sell candles that don't smell.

Because then it's 'percentage of reviews that complain there is no smell', removing the effect of an overall increase in reviews whatever they say.

That's assuming people that smell the candles just fine review at the same rate. I think we know that to not be true, so it's just burying the signal in noise.

If we categorized the reviews and the category of "no smell" increased in relation to the other categories, wouldn't that give a better view at the issue than looking at absolute values? Isn't this is what dividing by all reviews would effectively do, only in the context of a single category (no smell)?

Is it? IANA statistician and wouldn't claim it to be a strong suit, but isn't it only assuming that Smell and Not-smell each stay at the same rate (per sales)?

Then S/N is flat across the year (even as sales vary) iff opinion on smell (regardless of whether it leads to a review) is flat.

I've always called this the refrigerator problem. Reviews of major goods are allowed by the fact that people don't review them when they work as expected, only when they have issues. It's hard to imagine taking the time to review a refrigerator that reliably keeps my things cold. Some people do, but I think most don't.

Is it as misleading as calling the graph or chart a plot?

Seems a bit click-baity to me. it may be accurate but is exploiting the duality of the word "plot"'s meaning.

This ambiguity was not intended, I have edited my comment.

I meant the HN Title not Your comment, sorry for the confusion.

What absurd idea is it that makes you believe people are more likely to click on “plot” than “chart”? Nobody is reading that word an thinking “Gunpowder, Treason…”

Merry Christmas, thanks for keeping Your comment respectful.

>>>Nobody is reading that word an thinking “Gunpowder, Treason…”

Normally I wouldn't bother responding to a bully, but You are incorrect and I feel there are plenty of people other than myself that would read that in the way You discounted as foolish and unreasonable. Especially given the current times we are living in with a great pandemic that affects sense of smell, an abundance of conspiracy theories on any given subject and the name of the largest corporation in the world in the title.

You may never read a thing in such a way but to call others point of view as "Absurd" is "Arrogant" and "Conceited".

Happy New Years!

Reply on news.ycombinator.com