1
$\begingroup$

Given a dataset where 4500 reviews of consumer goods, provided via a 5 star rating system, have the following distribution:

  • 18% between 1 and 3 stars

  • 82% between 4 and 5 stars

Is it "right" to think that under similar conditions (i.e. same type and selection of goods) a binary rating system could yield similar results (18% thumbs down and 82% thumbs up) but with the advantage for the users to take less time to rate?

I assume that a binary rating system is quicker to use in comparison to 5 item-scale as mentioned here: Why would Netflix switch from its five-star rating system to a like/dislike system?

$\endgroup$
5
  • 1
    $\begingroup$ The only way to know whether a binary system takes more or less time to use is to collect data about user behavior and measure how long users spend rating items. This doesn't seem to be a statistical question. $\endgroup$
    – Sycorax
    Oct 20 at 16:05
  • $\begingroup$ Probably I didn't express myself correctly but I'm not interested in knowing whether the binary system is "faster" or not - that's a fact as per the link reported. I'm curious to know if the distribution of the results, that we would get using a binary system, is almost the same as the one with a 5 star system - ceteris paribus $\endgroup$
    – alz
    Oct 20 at 16:12
  • 1
    $\begingroup$ Why do you assume that 3 stars would be part of the lower category in a binary system? In fact you can't know this, and without any empirical research about it, nobody can. Furthermore, a binary system may attract some people that wouldn't rate using the 5 star system, and the other way round, so it would not be exactly the same people rating, so the distribution can also be different for that reason. (I'm not saying it will be, as there is no way to find it out except trying it out.) $\endgroup$ Oct 20 at 16:31
  • $\begingroup$ The unspoken question in your post is whether or not a partition of the quinary data is the same as if the data had been collected in binary form from the start -- this question is not answerable, because it's an empirical question about how users fill out forms. What problem are you trying to solve, and how is binning your data related to solving that problem? $\endgroup$
    – Sycorax
    Oct 20 at 16:38
  • $\begingroup$ Thanks for the feedback Christian Hennig and Sycorax $\endgroup$
    – alz
    Oct 20 at 17:15
0
$\begingroup$

Some people will only give a thumbs up if they view the product at a $5$. Some will for any rating of $4$ or higher. Some will for any rating higher than $1$. Some people will oppose such a rating system, preferring to be able to make a star-like rating. When you do the binning yourself, you get to make that choice.

I have my reservations about binning when you have a spectrum like you have, but I think the issues of human psychology will influence your data collection, yes.

$\endgroup$
1
  • $\begingroup$ Interesting take, thanks Dave! $\endgroup$
    – alz
    Oct 20 at 16:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.