# Thread: How should I explain this results?

1. ## How should I explain this results?

The hypothesis is:
the willing-to-pay is higher when people see print ad with testimonial.

So there are four print ads

1. 1 cola ad with testimonial
1. 2 cola ad without testimonial

I have
- a veritable named "adtype", 1 is "ad with testimonial", 2 is "without".
- another variable name "ad", 1 "cola with T", 2 "cola without T", 3 "shoe with T" and 4 "shoe without T".

After running a T-Test(group is "adtype", DV is "willing to pay"), I found out that there is:

a significant different between variable "willing-to-pay": ad with testimonial has a higher mean than ad without testimonial.

So far so clear.

But now I would like to look deeper in the data, so I run a one-way-ANOVA, ID is "ad", DV is "willing to pay", there comes my problem:

1. There is no significant different of "willing-to-pay" between "Cola with T" and "Cola without T". How does that happens?

2. What if there is no significant different between "Cola with T" and "Cola without T", neither a significant different between "Shoe with T" and "Shoe without T" (but there is a significant different based on the T-Test, group variable is "adtype"). What happens here?

I am quite confused now, hope someone would "inspire" me.

2. ## Re: How should I explain this results?

Try a 2-way ANOVA, including the interaction term.

3. ## Re: How should I explain this results?

hi,
you have a smaller sample size when you compare cola separately and shoes separately. The significance depends on the sample size as well, so you might have enough power to detect the difference with the aggregate test but not enough for the indiuvidual tests. The Simpson paradox could also be something to look at , but my guess is the loss of power.

BTW your variables are suboptimally defined - you could have a variable "Cola" or "Shoes" and another one "with T" and "without T" - that would describe the situation adequately and would be easier to analyse if you try a two way ANOVA.

regards

4. ## Re: How should I explain this results?

Originally Posted by rogojel
hi,
you have a smaller sample size when you compare cola separately and shoes separately. The significance depends on the sample size as well, so you might have enough power to detect the difference with the aggregate test but not enough for the indiuvidual tests. The Simpson paradox could also be something to look at , but my guess is the loss of power.

BTW your variables are suboptimally defined - you could have a variable "Cola" or "Shoes" and another one "with T" and "without T" - that would describe the situation adequately and would be easier to analyse if you try a two way ANOVA.

regards

Thank you rogojel so much for the detailed explaining. I would use "brand" and "adtype" variable and drop the variable "ad" with 4 labels.

Here I have another confuse about ancova.

I have another variable "how do you like the testimonials in the ad", and I assume that this variable is a moderate variable in my case.

Results from T test ( group "adtype", DV: "ad attitude"): significant different.

Results from Ancova (group "adtype", DV "ad attitude", Covariable "like of T" ):
-significant main effect of "like of T",
-no interaction.

Here I don't understand anymore, why does adtype no longer offer a significant difference?

Regards

5. ## Re: How should I explain this results?

Hi,
I am definitely not an expert in ANCOVA but I would guess that adtype and "like of T" are correlated. So, when the covariate was missing from the model the whole effect was attributed to the adtype, but if you take "like of T" into account it will sort of take over.

regards

6. ## The Following User Says Thank You to rogojel For This Useful Post:

yue86231 (11-24-2015)

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts