# Working with group totals...

#### HeatherHunni

##### New Member
Hi,

I was wondering does anyone have any thoughts on this.

I have recently conducted experiments looking at creative problem solving.

I had 30 participants, in groups of 5 assigned to one of three conditions.

I was looking at where whether brainstorming aloud or in silence was most effective so the conditions were Aloud, Silent, Control (participants brainstormed individually).

The dependent variable was the number of ideas produced by each group. So, I have scores for each of the six groups but not individual scores as the participants worked together.

I am also looking at a number of covariates.

Anybody any ideas on how I might analyse this?

Any help would be appreciated

#### Dason

So really you have six experimental units (the groups) and three treatments. Which means you only have two replicates of each treatment. The power to detect any differences is probably very small. Did you randomly assign the participants to groups and/or randomly assign the treatment to each group?

#### HeatherHunni

##### New Member

Yes, I randomly assigned the participants to the treatments..

#### Karabiner

##### TS Contributor
So you have n=6 (i.e. n=2 in each condition). If you want to do anything beyond descriptive statistics, you have to repeat the experiment with more groups.

Regards

K.

#### HeatherHunni

##### New Member
Hi,

No, I actually have 6 groups with five in each, 30 participants

Thanks!

#### Dason

So you have n=6 (i.e. n=2 in each condition). If you want to do anything beyond descriptive statistics, you have to repeat the experiment with more groups.
Realisitically they probably aren't going to find anything but it still is possible to find results. With a sample size of 2 you can do inferential statistics - they just won't be very powerful (or reliable).
Hi,

No, I actually have 6 groups with five in each, 30 participants

Thanks!
But you said you only have observations on the groups so really the groups are what you're observing correct? You only have six numbers to work with?

##### Ninja say what!?!
I don't know if this decision would be justified, but you could assume the same score for each person in the groups. How does the distribution of scores look for those who worked independently?

#### Dason

Could you explain how you assigned the treatments - what the treatments actually consisted of - how the data was collected in more depth?

For the "silent" group was there like... one sheet that everybody just wrote their ideas down on or did each person have a sheet and then the "group data" was just the compilation of all of this? For the group that got to speak was there just one sheet or something? How was the "control" treatment different than the other two treatments?

#### Karabiner

##### TS Contributor
No, I actually have 6 groups with five in each, 30 participants
Your unit of observation is group. You investigate how many ideas a GROUP produces
if a GROUP is instructed in this or that way. This understanding is not so much a matter
of statistics but of theory (I guess the idea that a group is something qualitatively different
from a bunch of individuals started in the 1920ies? or the 1520ies?). Fortunately, you do
not have observations on the individual level in the 4 real groups, so many possible false
analyses are prevented.

Regards

K.

Last edited: