0
$\begingroup$

I want to compare the job satisfaction of two groups with a two-sided t-test. My hypothesis is that the mean of group B is higher than the mean of group A

This are the results of my calculations:

  • Group A: mean = 3.79, SD = 0.99
  • Group B: mean = 4.07, SD = 0.90

t-value: -2.67** p-value= 0.0079**

I've used a the following code in R for the t-test

t.test(data$jobsatisfaction~data$group)

My question: How do I have to describe/interpret this result in correct terms for an academic paper?

(Do I have to use a one-sided-test? I am not sure, since my arguments for my hypothesis are kind of bad)

$\endgroup$
5
  • $\begingroup$ I do not think there is a simple answer to this question. Interpretations can be very different in different fields and settings. Sorry to say so, but probably the best way to know how to do it would be to read good papers in this field, or in another field, but describing similar data setting. $\endgroup$
    – cure
    Sep 17, 2020 at 18:58
  • $\begingroup$ How would you describe it? Just something like this: Group B has a statistically significant higher job satisfaction (p<0.01) than group A. $\endgroup$
    – user283542
    Sep 17, 2020 at 19:08
  • $\begingroup$ What are the conventions of your field and the journal to which you will be submitting? I would write something more like "Group B has a higher higher job satisfaction (two-sided t-test p=0.0079, delta=...)" but would go with however the publisher, professor, or boss wants it to go. // Very concerning, however, is your parenthetical comment at the end. In what ways are your arguments bad? $\endgroup$
    – Dave
    Sep 17, 2020 at 19:55
  • $\begingroup$ (Continued) Finally, if you want to use a one-sided test, that's fine, but you can't decide that after you see the difference and then test in that direction; you have to decide one-sided vs two-sided and the direction of one-sidedness before you see the difference. $\endgroup$
    – Dave
    Sep 17, 2020 at 19:55
  • $\begingroup$ This is my own research project in the field of human resource management. So would a two sided test be ok with those hypothesis' $\endgroup$
    – user283542
    Sep 17, 2020 at 20:05

1 Answer 1

0
$\begingroup$
  1. If you were before seeing the data interested only in a one-sided alternative like "mean of group B is larger than mean of group A" (for example because the situation is so that there is no good reason why it should be the other way round, or because you were specifically interested in something supposedly good that has been done to group B), you need a one-sided test. If you went into this open minded, thinking that anything may happen, and anything may be of interest, two-sided.

  2. I'd say "There's strong evidence in the data that job satisfaction in group B is on average higher than job satisfaction in group A." I wouldn't use "statistically significant" in the wording, because far more wars about the concept (of which there are many) are framed in terms of significance than in terms of strength of evidence.

  3. Note that the validity of this relies on model assumptions, random sampling of a sample representative of the population to which you want to generalise, and valid measurement.

$\endgroup$

Your Answer

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