1

I have used a survey to research willingness to pay regarding virtual goods in videogames. For this, I have asked the respondents a couple of questions via Likert scale, to get their opinions on certain statements. I then asked them if they had ever purchased virtual goods before.

In this case, I want to see if satisfaction with the game leads to a higher likeliness of purchasing virtual goods. How would I go about testing this statistically? My first idea was to convert my Likert data to numerical values, E. G. from strongly disagree = 1 to strongly agree = 5. However I can't seem to find which type of test I should use to then see if there is any correlation between a higher satisfaction value and buying virtual goods.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
R Vd H
  • 11
  • 1

2 Answers2

0

Strongly disagree to strongly agree is ordinal data. There is a large set of methods to use with ordinal data: spearman correlation or Kendall's tau may be the most prominent ones.

Bernhard
  • 7,419
  • 14
  • 36
0

You could use logistic regression. If you represent your likert, that is ordinal, predictor with numerical codes $1,2, \dotsc$, then use also a quadratic term to represent possible nonlinearity in the effects, or maybe even with a spline.

For an example with a monotone spline, see Looking for function to fit sigmoid-like curve. For some other ideas see How do you estimate correlations between ordinal and binary data in SPSS? and specifically Correlations between continuous and categorical (nominal) variables.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
  • The problem I'm struggling with is that when I look at any of the tests online, people are always using two numeric variables. How would I go about calculating the Spearman coefficient for example, with Yes/No data? – R Vd H Feb 12 '20 at 16:23
  • You shouldn't use Spearman, read also https://stats.stackexchange.com/questions/102778/correlations-between-continuous-and-categorical-nominal-variables – kjetil b halvorsen Feb 12 '20 at 17:17
  • The spearman correlation function in base R is corrected for this, no? – mribeirodantas Dec 22 '21 at 11:40