I have a question regarding mixed models. We are looking to describe physical activity in one group of children. Each child is wearing an accelerometer and will have 4-7 days of 'wear time', depending on how well they adhere to the protocol. For each child, we were going to estimate daily average physical activity (counts/15 seconds, a continuous variable). This is essentially a repeated-measures design, so the data is going to be arranged in long form (i.e, each child will have multiple rows of average daily physical activity).
The first objective of the study is simply to estimate average/SD physical activity in the sample. My first thoughts were just to generate an "overall average physical activity" measure for each child (e.g, for ID 1: OverallAvPA=(avPAday1 + avPAday2 + avPAday3 +avPAday4) /4), then to take the average of these individual averages to generate a sample average. However, it was suggested that we could instead use a mixed model without covariates to generate this summary statistic.
How would this be done? What is the benefit of this? I was under the impression that mixed models would only be used for looking at the relationship between variables, not to generate summary statistics.
The first objective of the study is simply to estimate average/SD physical activity in the sample. My first thoughts were just to generate an "overall average physical activity" measure for each child (e.g, for ID 1: OverallAvPA=(avPAday1 + avPAday2 + avPAday3 +avPAday4) /4), then to take the average of these individual averages to generate a sample average. However, it was suggested that we could instead use a mixed model without covariates to generate this summary statistic.
How would this be done? What is the benefit of this? I was under the impression that mixed models would only be used for looking at the relationship between variables, not to generate summary statistics.
Comment