Discover the world's scientific knowledge

Dear colleagues,

me and my research team are currently conducting a meta-analysis within the school context, so we face various studies that used multilevel analyses for their data. Therefore I have two questions:

1. Is there a "state of the art" method to address multilevel data within meta-analyses? E.g. if study one reported individual as well as class-level data and study two only reported class and school level data?

2. Is there a way to conduct multilevel analyses within meta-analyses? E.g. if all data was entered on individual level and if class size or number of classes is known, is there a way to cluster these data? I am aware that meta-analysis itself can be seen as some form of hierarchival analyses, but I am really interested if there is a way to analyze synthesized studies the same way primary studies analyze school data.

Thank you so much for your reply and best regards,

Angelika

Hi everyone, I'm preparing for an experiment and need to calculate the minimum sample size. Participants (level-2) are randomly assigned to one of the two conditions (level-1), and they will be measured with the same scales at two timepoints (level-3). Based on previous research, I decided that the effect size of condition on the outcomes is f=0.2.

Based on this design, how should I calculate the sample size I need?

Hi,

I currently have a dataset in which I would like model the temporal lag between my mediator variable and my outcome variable. Specifically, I have 4 time points for my mediator and my outcome variable and I am wondering how I go about analyzing this in SPSS. I want to understand how change over time in the mediator led to change over time in my outcome variable. 

I have three different scales in my analyses. Two are on a 6-point Likert scale (Independent variables), and one on a 5-point Likert scale (Dependent variable). For non-nested data, readings suggest standardizing the scale scores before analysis for comparison and interpretation.

For multilevel modelling, given group and/or grand mean centring the variables is recommended, do I still need to standardize prior to centring? Or will centring essentially equate to standardization of the variables?

In a daily diary survey, I have collected data for 10 days regarding 1 independent variable (IV), 1 mediator (Med), and 4 dependent variables (DV). I have also collected some more data before conducting the diary survey to use as moderators or control variables, for example, demographics-related (observed), motivation-related (latent variable).

This means the IV, Med and DVs are all Level 1 continuous variables, and moderators are Level 2 continuous and categorical variables.

In the beginning, I utilized MLmed macro for SPSS by Hayes and Rockwood (2020) to conduct multilevel moderated mediation analyses. However, I found many diary studies report the utilization of multi-level structural equational modeling (MSEM) in MPLUS. So I want to conduct MSEM/DSEM in Mplus because we can also check the model fit of our data, but I am not sure which example in the Mplus guide or which model and syntax would be the most appropriate for our data. The example of Mplus syntax provided in the Mplus guide regarding DSEM in chapter 9 (http://www.statmodel.com/ugexcerpts.shtml) does not contain any example 1-1-1 mediation moderated by level 2 variable in my understanding if anyone has any idea regarding Mplus syntax or resources for conducting 1-1-1 analyses moderated by level 2 variables that would be a great help.

I am trying to find an R code to conduct sensitivity analyses for minimum detectable effect for 1) linear mixed models and 2) multilevel binary logistic regression. I've looked into a few packages, including paramtest, simr, and PowerUpR, but have not been successful. Has anyone done a similar analysis or had luck with any of these packages or others? Thank you!

Investigate the effect of Condition (0, 1, 2, 3) on Inclusion per Demographic participant group (0, 1, 2, 3) over Time (0, 1).

I have been looking for procedures to calculate the sample size I need to be able to get 80% power to find an effect size of at least 0.2 (small according to Cohen) before data collection.

Using the book of Snijders and Bosker (1999) I found the following procedure:

1. Calculate Standard error (using Standard error calculation) with an effect size of 0.02 (small effect according to Cohen), an alpha of 0.05, and a power of .80. This lead to a standard error of 0.08. 2. Calculate n using the standard error from 1 (SE = .08), and assuming a standard deviation of 1 (and a mean of zero) under the normal distribution (using Standard error formula). This formula gives an n of 153.76.

3. Calculate the design effect (Design effect formula) with n = 153.76 and ρ = 0.08. This leads to a design effect of 13.22.

4. Multiplying n with the design effect in order to calculate the two-stage sample-size leads to a two-stage sample size of 2032

1. How can I know the intraclass correlation ρ before collecting data?

2. Why are there not parameters in the formulae about the amount of categories of my Condition, Demographic group, and Time variables?

3. Are there any easier procedures in order to calculate a sample size when using multilevel modeling?

Hi, I am performing multilevel analyses for individuals (level I), organizational characteristics in a hospital (II level) and outcome (e.g. survival). I have a confusion variable. How do I fit it into the model? (I'm using Stata). Many thanks!

Right now I am conducting a multilevel analyses with SPSS. I get outputs for random and fixed coefficients. What are the differences between them? How should I interpret both outputs?

Hi,

I am wondering how to get confidence intervals for coefficients in HLM software? I am using HLM8 specifically. It is curious that the program doesn't automatically report confidence intervals as SPSS does, especially because CIs are needed when writing up the results of multilevel analyses for papers. Have the developers decided not to include this feature or is there a way to get HLM to compute confidence intervals?