Help Plz! 2 way between groups anova.. Understanding results

Nowhaus

Hi
I hope someone can help me as I am getting quite distressed with spss and stats in general, I am useless with maths numbers stats etc (they are my secret gremlin that no one knows about ) and now I am trying to work out my hypothesis results in spss and I am Lost !! I don’t even know if I have found significant differences or if I have supported the hypothesis, I don’t think it has been supported but that is based on my own opinion and from looking at the survey results.
I have trawled the net looking for help with this but I’m getting nowhere quick and I don’t have allot of time, in fact I am quickly running out of it.
Below is a copy of the output results for a 2-way between-groups ANOVA.
The two hypothesis are that females will be more persuaded to purchase online P.R AND
Secondly that those with low need for cognition scores (scored, 1 for high Need for Cog and 2 for low Need for Cog) will be more persuaded to purchase online P.R
From what I have gathered... the first two boxes of information are more or less superfluous and the results lie in the third box of information, so I have only pasted the third box below.
I would really appreciate if anyone can help me out here by explaining where the significant results are and how I report that in my paper, I’m sure it’s probably very easy and probably a really stupid obvious question, but I have no experience with SPSS and this is like another language to me.
Thanks so much!
All feedback is greatly appreciated!!!
so confuddled, im pulling my hair out here!!

Tests of Between-Subjects Effects
Dependent Variable:Have you ever purchased a PR. suggested for you?
Source Type III Sum of Squares df Mean Square F Sig. Partial Eta Squared
Corrected Model 1.123a 3 .374 1.558 .214 .102
Intercept 226.909 1 226.909 944.059 .000 .958
Sex .070 1 .070 .290 .593 .007
NeedforCog .740 1 .740 3.077 .087 .070
Sex * NeedforCog .453 1 .453 1.883 .177 .044
Error 9.855 41 .240
Total 310.000 45
Corrected Total 10.978 44
a. R Squared = .102 (Adjusted R Squared = .037)

2Scoops

Nowhaus wrote: »

The two hypothesis are that females will be more persuaded to purchase online P.R

Secondly that those with low need for cognition scores (scored, 1 for high Need for Cog and 2 for low Need for Cog) will be more persuaded to purchase online P.R

These are two separate hypotheses. Two 1-way ANOVAs would do the job. A 2-way will also detect the interaction effect, if any, but as you haven't hypothesized such, it's irrelevant.

Nowhaus wrote: »

Tests of Between-Subjects Effects
Dependent Variable:Have you ever purchased a PR. suggested for you?
Source Type III Sum of Squares df Mean Square F Sig. Partial Eta Squared
Corrected Model 1.123a 3 .374 1.558 .214 .102
Intercept 226.909 1 226.909 944.059 .000 .958
Sex .070 1 .070 .290 .593 .007
NeedforCog .740 1 .740 3.077 .087 .070
Sex * NeedforCog .453 1 .453 1.883 .177 .044
Error 9.855 41 .240
Total 310.000 45
Corrected Total 10.978 44
a. R Squared = .102 (Adjusted R Squared = .037)

It's hard to read in this without the correct formatting. Look for the column that lines up with 'Sig.' - this is the significance of your F-statistics (the ANOVA) and will tell you if any Sex or NeedforCog differences are 'real' or just a coincidence. I think I have highlighted the numbers that line up with Sig. in red, but maybe I'm reading it wrong.

Assuming your a priori alpha value is 0.05, it looks like that Sex does not influence online P.R buying (whatever that is!). NeedforCog exhibits a non-significant trend (is your study adequately powered?) and there is no interaction effect i.e. both sexes look similar within both levels of NeedforCog.

Nowhaus

Thanks so much for the help. So you are suggesting I use two, 1 way between subjects? anova? My supervisor had advised me to use the 2 way between groups anova.
How would I report the results in my paper, is there a certian way you would suggest or a site that can help expalin how to report them? I have SPSS for psychologists by N. Brace, R.Kemp & R. Snelgar to try help me with this and it makes some suggestions on how to report these results, but I cant follow and apply them to my results (because I dont understand where some of their points/figures come from in their examples... so I cant attempt to repeat the report with my results)

Oh P.R is personalised Reccomendation, for example when you shop on Amazon and it suggests a product you may wish to purchase based on your previous purchase.


The whole thing is an awful mess.... I cant understand the spss output and have spent well over a month trying to understand stats and spss but as i mentioned numbers are my weak point so its been a loosing battle. Thanks appreciate the help


p.s

I have a correlation and mann whitney I am trying to work out... If I have no luck they may appear up as another question in the next while....

Thanks again, I Really appreciate your help!

2Scoops

Nowhaus wrote: »

So you are suggesting I use two, 1 way between subjects? anova? My supervisor had advised me to use the 2 way between groups anova.

You only need a 2-way if you're asking a question related to the interaction. If that's not your question (and it's not, based on your 1st post in this thread), then it's irrelevant. However, do what your supervisor says, because they're in charge and you don't want to annoy them.

Nowhaus wrote: »

How would I report the results in my paper, is there a certian way you would suggest or a site that can help expalin how to report them?

There are many ways to report these kind of data and personal preference will probably determine how you go about it. Check with your supervisor. You may do something as simple as 'men were 40% less likely to buy PR online (P<0.05) compared with men" or your supervisor may want all sorts of regression coefficient data in there instead. My advice is to keep things as simple and coherent as possible, while providing the advanced data to those interested. And include graphs, where appropriate.

Nowhaus wrote: »

The whole thing is an awful mess.... I cant understand the spss output and have spent well over a month trying to understand stats and spss but as i mentioned numbers are my weak point so its been a loosing battle.

For a simple analysis like this, it's not key that you understand every single bit of SPSS output, but rather the ability to extract the needed information for your results. Find out what you need to know (the effects of sex etc), graph the data so you can see the differences (if any), then check the SPSS output to see if they have a statistically significant influence.

Nowhaus wrote: »

I have a correlation and mann whitney I am trying to work out... If I have no luck they may appear up as another question in the next while....

No worries - two more tests that can be quickly explained. :pac: If it makes you feel any better, it's your supervisor's fault for not preparing you properly.

Nowhaus

Thank you so much for your help, I really do appreciate it... I dont think I can really blame my supervisor for my lack of spss knowledge because i missed a few lectures.and had to cancel some meetings due to family illness and then bereavement, so I totally acknowledge my own doing in getting into this mess.... (this is part time study and I work full time so catch up is extremely hard) but I did kind of expect a bit more help from her with the spss than was offered... perhaps I had very high expectations of how much assisstance I would get in this area, seen as we had so few lectures on stats or spss. ANYWAY.......


Quick Question relatin to my last two hypothesis if you dont mind. My supervisor has suggested that I use Pearsons Correlation for hypothesis 3... but im not so sure that this is the correct test to use....
Hypothesis three.
Females are also more likely to participate in the interactive elements of Amazon like communities and reviewing and rating the products.

So here I am looking at gender and a question where the participants were allowed to pick multiple choice of the interactive elements they use on Amazon ... for example, perhaps they like to rate products, take part in the community and create and interact with other users wish lists... so they could pick all three options if they wish... or answer I do not use any interactive elements on Amazon.....
My supervisor originally suggested that this and hypothesis 4
(Hypothesis 4: It is hypothesised that users with a low need for cognition are more likely to participate in the interactive elements of the site like communities and reviewing and rating the products)... were ranking questions, and then she either got confused or changed her mind and two minutes later advised I do a correlation and a mann whitney for Hypothesis 4.... I think she must have made a mistake... because when I look at them now I think that maybe they should be ranking.(I did ask at the time and she confirmed Correlation for H3 and Mann whitney for H4)

Oh and regarding these questions being ranking..... very few survey participants actually took part in any interactive elements on Amazon... so I was thinking (because I am unsure if I have a multiple response question correctly inputted into SPSS... and because this is all very confusing) would I be better off just inputting participants as 1, use interactive elements and 2 do not use interactive elements and then conduct whatever test... correlation for example....ranking seems very complicated and the books I have seem to refer to ranking in order of small to large and my question is not like that at all.. its multiple choice.
Then I can assess the descritive results of what percentage of users, use what interactive elements etc... seperatly.
Sooooo.... any ideas, suggestions... as you can see Im not so sure my supervisor is all that knowledgeable of SPSS.:rolleyes:.. I know she was unsure with some areas....
sorry did I say a quick question..... a long winded one is what I actually meant.
Thanks again for the help!!

2Scoops

Nowhaus wrote: »

My supervisor has suggested that I use Pearsons Correlation for hypothesis 3... but im not so sure that this is the correct test to use....

Without getting into too much detail, and possibly adding more confusion, there is a degree overlap with correlation/regression and ANOVA/t-test whereby they ultimately arrive at the same statistic, albeit by different routes. So, a simple correlation like Pearson will indicate if there is an influence of sex, for example. However, since sex here is a discrete group, rather than a continuum, a t-test/ANOVA would yield the same result but via a more appropriate method.

However, it's a little more complicated than that, in this case. A lot depends on how you express the 'engaging in interactive elements' variable. You could do it as Yes or No - that would be a rank: Yes is more than No, but it is unclear by how much. Or, you could have it as 'number of elements interacted with:' 0, 1, 2 etc. I think it would be best to treat this as a rank as well, since I think it might not be correct to assume that all interactive elements are equal. Anyway, this would indicate a non-parametric approach, and Pearson's would not be appropriate anymore. Mann-Whitney or chi-square can handle it, either way. Spearman's rank test will get you a correlation coefficient, if you need one.

Same goes for hypothesis 4.

Nowhaus

Thanks so much for all your help!! I really appreciate it! Thanks!:)