Help with SPSS test.

#1
Hi All,

NOOB here! I am trying to decide what tests to run on the data from this experiment with 30 participants;

PART 1
We recorded how much television (in minutes) each participant watches in day, and how much of this time was spent watching cartoons (in minutes), they then complete a psychometric test and receive a score between 0 -21.

PART 2 We recorded how much television (in minutes) each participant watches the next day, and tell them to try and abstain from watching cartoons, they then complete the same psychometric test and receive another score between 0-21.

example data returned
PART1
Participant ID 1001 - watched 270 minutes TV in total, and 70 minutes cartoons and scored 14

PART 2
Participant ID 1001 - watched 200 minutes TV in total, and 5 minutes cartoons and scored 12

All good up to this point. I have run checks that data is normally distributed and the paired t-test shows p<0.026.

My issue is what test to run after this? I want to analyse how the number of minutes in total and number of minutes on cartoons has an effect on the score and any interplay.

My first instinct was a multiple regression then a repeated measure anova? what do you think?

Thanks,
Jim
 

Karabiner

TS Contributor
#2
That looks like a case for multilevel modeling. The 2 time points are nested within participants.

With kind regards

Karabiner
 
Last edited:
#3
Thank you, I'm not familiar with this analysis so I will have to do some research on it...

For the analysis would it be better if I enter the times in SPSS like so?

Participant ID 1001 - 200 minutes TV excluding cartoons, 70 minutes cartoons and scored 14.

instead of;

Participant ID 1001 - 270 minutes TV in total, and 70 minutes cartoons and scored 14.

or would the second version allow more analysis of how watching cartoons effects total time watching TV? e.g. it may increase total time and this may affect the overall score.

My hypothesis is that that watching cartoons correlates with an increase in the psychometric score and reducing the number of minutes watching cartoons will decrease the score on the psychometric test.

Hope that makes sense.

Jim
 
#4
I have decided to go with this method

Participant ID 1001 - 200 minutes TV excluding cartoons, 70 minutes cartoons and scored 14.

and keep it simple by just concentrating on the effect the cartoon variable has on the psychometric score. Would a multiple regression be enough for that?
 

Karabiner

TS Contributor
#5
In a multiple regression, each participant would appear twice, and this would
violate the assumed independence of observations.

Your first hypothesis could be analysed by two cross-sectional linear regression
analyses (one for t1 and one for t2). The second one would be a correlation between
the differences (minutes for watching cartoons at t2 minus minutes for watching
cartoons at t1, versus: scor at t2 minus score at t1).

With kind regards

Karabiner
 

Karabiner

TS Contributor
#10
With this I meant:
if you do not consider the longitudinal aspect of the design at this point,
but rather consider it as 2 cross-sectional surveys.
 
#11
With this I meant:
if you do not consider the longitudinal aspect of the design at this point,
but rather consider it as 2 cross-sectional surveys.
Ok, I think I understand. I will ignore the longitudinal aspect at present. I'm going to execute the multiple regression analysis, will be add my output to make sure we are on the same page :)