[color=#000000]Hi Natasha,[/color]
[color=#000000]That's all perfectly fine, and scores with "actual" values of 0 are treated appropriately in group-specific covariates (and differently from the subjects in other groups even if they share those "0" values). Relatedly, coding with "0" the subjects in the opposite group in these group-specific covariates is often just a matter of simplicity/convention, and the choice is irrelevant for most analyses. For example, in your Drug,Placebo,ScoresDrug,ScoresPlacebo [0 0 1 -1] analysis, the results will be [b][i]exactly [/i][/b]the same if you code your Placebo subjects in the "ScoresDrug" covariate and your Drug subjects in the "ScoresPlacebo" covariate with a value of 100 (or any other arbitrary value) instead of 0. Yet I still recommend coding those "opposite-group" subjects with 0's instead of any other arbitrary value, just because in the VERY few cases where that choice matters the choice of 0 is almost invariably the appropriate one. [/color]
[color=#000000]Hope this helps[/color]
[color=#000000]Alfonso[/color]
[color=#000000]
[/color][i]Originally posted by Natasha Mason:[/i][quote]hello all,
I've been running an analysis, where I want to look at the association between scores on a test and connectivity in two different treatment groups (drug vs placebo; between subjects). I have been running the analysis according to previous help I found on this forum: https://www.nitrc.org/forum/message.php?msg_id=14045
Where in the second-level analysis i've 1) created 2 second level covariates (drug, placebo) containing 1s and 0s.
2) created my test scores in the second-level covariates; an overall test scores with all scores in it ('scores'), and 2 additional covariates divided by groups (ScoresDrug, ScoresPlacebo; where I put a 0 for the everyone who is not in that group).
I then select Drug, Placebo, and Scores and enter a contrast of [0 0 1] to look at association between symptom scores and connectivity across all of your subjects (jointly across the three groups) after discounting potential differences in average connectivity between your groups.
I also select Drug, Placebo, ScoresDrug, ScoresPlacebo and enter a contrast of [0 0 1 -1] to look at the difference between drug and placebo in their association between scores and connectivity.
My question, however, is that for some of these participants, the actual scores are a "0", which shows that there is no difference from the baseline test we did with each participant. Thus I'm wondering if I am losing this information, as I code participants as 0's in the treatment specific contrast to split them into groups. They can also score negatively on the test, so I cannot just add a constant to each participant.[/quote]
[color=#000000]That's all perfectly fine, and scores with "actual" values of 0 are treated appropriately in group-specific covariates (and differently from the subjects in other groups even if they share those "0" values). Relatedly, coding with "0" the subjects in the opposite group in these group-specific covariates is often just a matter of simplicity/convention, and the choice is irrelevant for most analyses. For example, in your Drug,Placebo,ScoresDrug,ScoresPlacebo [0 0 1 -1] analysis, the results will be [b][i]exactly [/i][/b]the same if you code your Placebo subjects in the "ScoresDrug" covariate and your Drug subjects in the "ScoresPlacebo" covariate with a value of 100 (or any other arbitrary value) instead of 0. Yet I still recommend coding those "opposite-group" subjects with 0's instead of any other arbitrary value, just because in the VERY few cases where that choice matters the choice of 0 is almost invariably the appropriate one. [/color]
[color=#000000]Hope this helps[/color]
[color=#000000]Alfonso[/color]
[color=#000000]
[/color][i]Originally posted by Natasha Mason:[/i][quote]hello all,
I've been running an analysis, where I want to look at the association between scores on a test and connectivity in two different treatment groups (drug vs placebo; between subjects). I have been running the analysis according to previous help I found on this forum: https://www.nitrc.org/forum/message.php?msg_id=14045
Where in the second-level analysis i've 1) created 2 second level covariates (drug, placebo) containing 1s and 0s.
2) created my test scores in the second-level covariates; an overall test scores with all scores in it ('scores'), and 2 additional covariates divided by groups (ScoresDrug, ScoresPlacebo; where I put a 0 for the everyone who is not in that group).
I then select Drug, Placebo, and Scores and enter a contrast of [0 0 1] to look at association between symptom scores and connectivity across all of your subjects (jointly across the three groups) after discounting potential differences in average connectivity between your groups.
I also select Drug, Placebo, ScoresDrug, ScoresPlacebo and enter a contrast of [0 0 1 -1] to look at the difference between drug and placebo in their association between scores and connectivity.
My question, however, is that for some of these participants, the actual scores are a "0", which shows that there is no difference from the baseline test we did with each participant. Thus I'm wondering if I am losing this information, as I code participants as 0's in the treatment specific contrast to split them into groups. They can also score negatively on the test, so I cannot just add a constant to each participant.[/quote]