I conducted an experiment, in which I measured participants altruism (and scored them from 1 to 5 according to their answers means) and let them indicate their willingness to pay for multiple products. Additionally I surveyed their gender, age, income and level of education. Using pearson correlations, i can tell that altruism indeed had an effect on the willingness to pay for certain products.
My question now, how do I check for effects of the socidemographics in this relationship between altruism and WTP?
My university's supervisor suggested to run linear regressions with the WTP as dependent variable and altruism and the sociodemographic variables as independent variables. But instead of simply plugging in the education variable (that has a value for each level of education) i should rather compute new dummy variables. For education for example that would be a dummy variable that has value 1 if the participant has a bachelor's degree and 0 otherwise, another variable that is 1 if the participant has a master's degree and 0 otherwise. I don't really get why I should do this for the regression. Wouldn't it make more sense to do t-tests in this case?
Maybe I completely misunderstood him but I do not understand why I should do it in this way. Maybe someone can clarify that for me.
Thanks!
Hi, thanks for your quick reply. Unfortunately the slides are in albanian (i believe) which I do not speak.
I measured age continuously (participants entered their age in years). I do not know what you mean by "pull it out and split the results by gender". So you mean I should separate my data in male and female and run regressions for those two "new" datasets? How would I then interpret the results?
For the approach of my supervisor: I measured education as a categorial variable (1 = no finished education; 2 = high school degree; 3 = bachelor's degree; and so on). If I now create dummy variables for each manifesto of the education variable, do I simply run multiple linear regression and plug in one of the dummy variables for education every time?
I would like to use the linear regression as my supervisor suggested to use it or would you say that GLM is better for that case?
So I just ran the regression with my main effect "altruism" - "wtp", controlling for gender, income and education. For each manifestation of those categorial variables I created a dummy variable (income was measured in 5 levels and education was measured in 6 levels). The variable "female" is 1 for females and 0 for males. I left out edu_1 and income_1 to use them as baseline.
The Syntax:
REGRESSION
/MISSING LISTWISE
/STATISTICS COEFF OUTS R ANOVA
/CRITERIA=PIN(.05) POUT(.10)
/NOORIGIN
/DEPENDENT wtp_utz_1
/METHOD=ENTER altruism female income_2 income_3 income_4 income_5 edu_2 edu_3 edu_4 edu_5 edu_6.
If you have look at the output what would the interpretation be? I can't help myself and am really confused. Does it mean, that gender significantly changes the effect but income and education have close to no influence on the relationship "altruism" - "wtp"?
Kind regards
The frequencies are as follows:
edu_1 = 5
edu_2 = 36
edu_3 = 39
edu_4 = 58
edu_5 = 8
edu_6 = 3
income_1 = 14
income_2 = 47
income_3 = 26
income_4 = 17
income_5 = 35
9 did not want to indicate their income
Gender is actually very skewed with 120 females and 29 males.
I am quite aware that this is not optimal and that I should be careful with the results, especially with those for education.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression