Good afternoon,
I am working on my thesis on team work and I need to compare if teams differ when I adopt the rule 'at least 2 members of the team gave valid data' (19 teams in total), against the rule' at least 50% of the team members answered (17 teams in total)'. The team size is sometimes 12 members, so In order to take in consideration more teams for the research, I need to check this. If there is no significant difference in teams if both rules are applied in both rules, I would have more teams for the research if I adopt '>2 team members' rule.
I did add new data set with dummy variable, >50% vs >2 people responded. In that way most subjects overlap in the same data set. With this results I did not get significant difference.
But how justified is this analysis because it will be like repeated measures on independent T test? (I hope it makes sense:)
Or is there is another way to test this in SPSS?
Thank you all in advance,
Kindly regards,
Sonja V.
Hello Sonja,
So you are classifying the same batch of teams multiple times, applying a different criterion, right?
It wasn't clear whether you were more interested in: (a) how many teams met each criterion; (b) some difference (on some unnamed dependent variable) between teams that did meet vs. teams that did not meet some specific criterion.
If the point of interest is (a), then you could use the McNemar test, looking at the instances in which criterion A and criterion B differentially classify a team (e.g., the disagreements). However, with the number of teams you apparently have, this will not be a very powerful test.
If the point of interest is (b), then you'd be running multiple t-tests (or equivalent), each one based on a different criterion.
You'll have to decide which (if either) is more salient to your research question(s).
Good luck with your work.
Dear Mr. Morse,
Thank you for your time and willingness to help me solve this issue. It is matter of overlapping samples. I red afterwards that I did not explained the issue really good.
Lets say I have 20 teams for the research, predictor, moderator and DV. Initially the idea was to do the research with teams with response rate more than 50%. But as the data collection was a big struggle, e.g. I got big teams of 12 people and 5 of them only completed the survey, so I thought I change the criterion and instead '>50%', to make it at least 2 team members gave valid data. Off course, I need to check if this is valid decision, I.e. to check if there is significant difference on all the variables involved in the research if I work with the 17 teams ">50%response rate" vs 19 teams (when I adopt the criterion >2 team members gave valid data).
What I did do far is I add in data base all teams (19+17), even that most of the teams overlaps, and that way I could make independent T test (with dummy coding, ">50%' and ">2' members, which was not significant on any of the variable. But I am afraid that this way the true "error" around the means cannot be reflected?!
Thank you in advance,
Kindly regards,
Sonja V.
Hello Sonja,
Yes, it can be hard to collect data!
From your second description, it sounds as if you are comparing 17 teams to themselves + 2 others. Those will not be independent cases, so the usual independent t-test would not apply. If the two criteria yield 17 of the same batches and differ on only 2 others, I don't think that the difference between groups would be noteworthy, no matter how you compared them.
If the proximate goal is to have as much usable data as possible (without having to go and collect more), then though the less stringent two or more per group nets you two more batches, I'd stick with the 50% threshold, since you would have an easier time defending that as likely yielding a response set that could be representative of the entire group (and that the less stringent criterion only would have yielded two more groups).
Good luck with your study.
Dear Mr. Morse,
Thank you very much for your answer. I really appreciate your advice. In such a case when I need teams as much as possible, that two extra would be high appreciated.
I though so that the teams would not be independent cases...The problem is that it is not the same batch of teams on both criteria. Different teams met different criteria. I sent you a scheme of which teams are not included in the samples according to each criteria. Maybe it will yield some other ideas.
Thank you in advance,
Kindly regards,
Sonja
Hello Sonja,
From your work page, note that: (a) Group 21 appears only on the "> 2" side of the display, but not on the "50%" side; (b) the "> 2" should be ">= 2".
As is, (and discounting group 21 because it appears only once), you have:
15 duplicated sets (14 groups met both criteria; 1 group met neither)
2 sets that met the 50% but did not meet the at least 2 criterion
4 sets that met the at least 2 criterion but did not meet the 50% criterion
So, indeed, the groups are pretty much overlapping (15 of 21). All you'd be able to do with independent groups is compare the 2 sets (meeting only 50%) to the 4 sets (meeting only at least 2 criterion), but that seems somewhat futile, given the small numbers of groups.
The only scenario I can envision that allows a "50%" case to be met, but not the "two or more members" case to be met is that you have variable group sizes, and at least one had only two (2) members.
What you should probably do, if that is indeed the case, and your goal is to maximize the number of groups for which you can use data is modify your rule to:
"50% of the group OR at least 2 members"
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS