I am writing my thesis and am stuck on the data-analysis part. I want to find out how having a reservation influences revenue.

I want to find out whether guests with a reservation spend more time at the table than walk-in customers, and whether they spend more money. After that I would like to find out whether the day of the week has an influence on either of these findings (weekday or weekend), whether time of eating should be taken into account (lunch or dinner) and whether being a first time visitor or returning influences the outcome.

I have a large dataset (8240), cleaned up and ready to be used. It includes:

  • Minutes spend at the table
  • Spending per person
  • Dummy variable whether lunch (0) or dinner (1)
  • Dummy variable whether walk-in (0) or reservation (1)
  • Dummy variable whether single (0) or returning visitor (1)

Any suggestions/ideas on how to handle this? I have done a multiple regression analysis but am not sure how to include the different factors correctly. I can use Excel or R.

Distribution of spending

Output regression R

  • $\begingroup$ Please post your output $\endgroup$
    – user10619
    Jun 18, 2017 at 9:44
  • 1
    $\begingroup$ What you have done seems a good start. If you are using R then declare your three predictors as factors and R will take care of things for you. $\endgroup$
    – mdewey
    Jun 18, 2017 at 10:44
  • $\begingroup$ Be sure to plot and analyze the residuals $\endgroup$ Jun 18, 2017 at 11:07
  • $\begingroup$ I have added the output I got after declaring the three as factors $\endgroup$
    – Emily
    Jun 18, 2017 at 13:33
  • 1
    $\begingroup$ plot(model1) will generate a series of useful diagnostic plots. Those plots include a lot of information, so you might contact your university's statistics department to see if you can get a tutor or consultation for an hour or so. Because of that long right tail, pay special attention to Cook's distance (see this nice answer here: stats.stackexchange.com/a/206330/68397). Depending on your exact research question, cross validation or robust regression (stats.stackexchange.com/a/104371/68397) might be worthwhile. $\endgroup$
    – Dan Hicks
    Jun 18, 2017 at 14:21

1 Answer 1


To define the type of stat model you need to understand the behaviour of your dependent variable. In your case, it seems that your DV ("revenue") is at least continuous. Therefore you can forget about logistic regression, which is suitable for binary (0/1) dependent variable. Difficult to advise on the exact type of model you need without more info re your DV - Look at the distribution (Is it normally distributed?); Is the variable truncated? etc.

  • $\begingroup$ Thanks for thinking with me! I have added a picture of the distribution of my DV, seems to be quite normally distributed with a few outliers (should I remove them?) $\endgroup$
    – Emily
    Jun 18, 2017 at 13:02
  • $\begingroup$ The first thing to do is to double check them for errors in data entry. The second thing is to check them for plausibility (can someone really spend 5000 of your currency units on a meal when most people spend 500? The take @DanHicks valuable advice in his comment on your question to see why they are where they are. Do not delete them until you are sure that is justified and, of course, report what you did in your write-up. $\endgroup$
    – mdewey
    Jun 18, 2017 at 15:36
  • $\begingroup$ You do have some outliers in your database - Eventually you could use log() transfo to improve the normality of the "revenue" variable. Then a "classical" linear regression seems appropriate in your case. Not sure what the objective of your analysis, but the "spending per person" variable would probably require extra attention - Might be that the assumption of independence is not fully verified in your case $\endgroup$
    – Nicolas K
    Jun 18, 2017 at 20:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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