Modeling Question in a Daily Diary Study?

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

My Question is  how to use HLM in a daily diary study to answer the research question.

The study design is as follows:

14 days of measurement to capture within-person variability in mood, impulsive behaviors, etc (1 measure per day)

2 cortisol samples are taken each day, at waking and then 30 min later(2 measures per day - actually there are 4 since each sample is split and analysed separately for reliability. I use these as split scales so I can model the linear change in cort from waking to 30 min later. this parameter is of particular interest to the PI on the study and is the one I'm having trouble with). My baseline Level 1 model is

Cortij = B0j + B1jTime + B2jcDay7 + rij

(The B2 term is just to control for any potential linear effect of days per Bolger and Laurenceau-  I account for it but it's non-significant)

The research questions relate to how within-person fluctuations in mood/impulsive behaviors predict BOTH waking cort level and the slope from waking to 30 min later (Time coded 0 1). As an example: "When a person drinks more than their personal average, is there an associated increase in waking cortisol, or a steeper increase in cortisol from waking to 30 post waking?"

I thought putting in the person-centered predictor (in this example drinks), and then an interaction of the person centered drinks variable with time at Level 1, might give me my answers, but a colleague has indicated this is not correct. Any modeling suggestions to answer the above design question are most appreciated.

Thank you,

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：question Modeling Quest study DIARY software research actually familiar measures

I know nothing about cortisol, and I think that it is the base for your modeling that there is a relationship between time and the level of cortisol.In the second place, the arrangement of your variables seems to be reasonable, however, the way you center the predictors seems to be suspectable. In general, grand-mean centering is used.