# Which hotel is better?

#### TeoK

##### New Member
Hi everyone.
Maybe this is stupid, but when we chose a hotel for a holiday on TripAdvisor and Bucking.com, I had an idea about how to compare two hotels by their rating(despite the fact that there is already a general rating). For example, TripAdvisor: Grades are divided into 5 categories (Ordinal), from best to worst. And there is a general rating from one to five, taking into account the assessment in each category and the number of people who gave the assessment. For example, in hotel A, which rated 2297 in total and respectively:
Excellent 1166, Very good 718, Average 273, Poor 89, Terrible 51. Overall rating 4 out of 5.
Hotel B, totaling 4349 and respectively:
Excellent 3553, Very good 582, Average 135, Poor 49, Terrible 30. Overall rating 4.5 out of 5.
Having done the analysis in the spss, anova except that the two hotels of the Significant are different, I did not achieve anything.
Is comparison and statistical analysis possible? After all, in hotel A, it was "visited" 2297, and in hotel B, "it was 4349, as it seems to me, the overall rating may not correspond to reality.
My question is, is it possible to make an analysis based on assessments, could anyone have already tried to conduct an analysis? or can add more data (for example price)? Or am I just waste time?

#### CamilleJosion

##### CaJosion
Hi Teok. This is a typical case where you can summarize your data in a contingency table

You can imagine that each of the 6646 individuals tells which hotel he went to and gives a rating. You can then run a Chi-square test to make sure there is independence. Meaning that what you observe is not likely to happen if people would be answering questions at random.

I don't use SPSS but XLSTAT, SAS and R. Here are the results with XLSTAT

So there is definitely a structure. Now, I believe you would like to choose a hotel rated very good or excellent. So I merged the cells into

and ran a Fisher's exact test and there again

So you would definitely go for hotel B.

However, something comes to my mind. The individuals should be "independent" themselves. It is not perfectly true, as the previous ratings influence which hotel people will choose, but it is reasonable to assume their rating is not influenced.

#### Attachments

• 2.3 KB Views: 3
Last edited:

#### Durandot

##### New Member
Hi Teok. This is a typical case where you can summarize your data in a contingency table

View attachment 1249

You can imagine that each of the 6646 individuals tells which hotel he went to and gives a rating. You can then run a Chi-square test to make sure there is independence. Meaning that what you observe is not likely to happen if people would be answering questions at random.

I don't use SPSS but XLSTAT, SAS and R. Here are the results with XLSTAT

View attachment 1250

So there is definitely a structure. Now, I believe you would like to choose a hotel rated very good or excellent. So I merged the cells into

View attachment 1252

and ran a Fisher's exact test and there again

View attachment 1253

So you would definitely go for hotel B.

However, something comes to my mind. The individuals should be "independent" themselves. It is not perfectly true, as the previous ratings influence which hotel people will choose, but it is reasonable to assume their rating is not influenced.
Sorry to bring this up, but thank you for sharing, it was helpful to me. I have read your post quite a few times and I liked the way you were thinking. That's the way I will choose my hotel next time. Last year, we were advised by friends to choose a hotel on TripAdvisor. I can confidently say that the rating of the hotel (can't remember the name of the hotel) was 4.5 out of 5. I strongly disagree with this, I would have given 2 points at all, just for the comfortable bed. I would have given 2 points just for the comfortable bed. We spent more time in the clubs than in the hotels.

Last edited:

#### gw1

##### Member
Are people less likely to rate 1 or 2 unless it's really bad? So a test to avoid the one with relatively more poor and terrible scores might be interesting