Timeline for HELP: LM shows no relationship, but LMM does
Current License: CC BY-SA 4.0
18 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 6, 2019 at 23:53 | history | edited | mucaua | CC BY-SA 4.0 |
added 15 characters in body
|
| Aug 6, 2019 at 14:27 | comment | added | mucaua | To clarify -- My question is whether the LMM (#3) is able to find relationships between x and y when considering and accounting for data for all populations together, even though LMs for each population separately cannot. I understand that it is. CloseToC indicates that it is. I ask this because a colleague suggested the results of the LMM (#3) were misleading due to Simpson’s Paradox. | |
| Aug 6, 2019 at 14:27 | comment | added | mucaua | Sorry if my question is confusing. I’ll try to clarify a few things: - R results in #1 show LM fitted individually to one population, and the graph is for all populations, each of which fitted individually. - Results and graph in #2 show LM fitted to all data from all populations, without accounting for population-level effects. - Results and graph in #3 show LMM fitted to all data for all populations but with populations accounted for as random effect - Data used in all analysis come from the same data frame. | |
| Aug 5, 2019 at 14:10 | comment | added | mkt | @user158565 See the text above the first block "See one example of my preliminary LM results for one such population below." | |
| Aug 5, 2019 at 14:09 | comment | added | mkt |
@user158565 df is being redefined in between the two blocks of code (not shown). That's my guess at any rate. Not suggesting it's a good idea.
|
|
| Aug 5, 2019 at 14:02 | comment | added | user158565 |
@mkt But both of them have data=df.
|
|
| Aug 5, 2019 at 8:51 | comment | added | mkt | @user158565 The code is the same, but the dataset is different - note the difference in the degrees of freedom in the model output. The first dataset is one group, the second is the full dataset with many groups. I agree that it is confusing. | |
| Aug 4, 2019 at 0:17 | comment | added | user158565 |
Anyone find the difference between first two call: lm? If not, why the same code generated the different results?
|
|
| Aug 3, 2019 at 22:37 | comment | added | CloseToC | Simpsons Paradox in this case is the simple fact that x is approximately useless for predicting y within populations, but is useful for predicting y when you pool data, because the population averages of x and y are correlated. The linear mixed model is a middle ground between estimating a linear model population by population and estimating an overall model. It will not give substantially different results than these two results you already have | |
| Aug 3, 2019 at 22:00 | review | Close votes | |||
| Aug 4, 2019 at 11:13 | |||||
| Aug 3, 2019 at 21:47 | comment | added | mkt | I agree with Isebella Ghement and don't understand what problem this procedure solves. It would help if you clarified this with more specifics. | |
| Aug 3, 2019 at 21:17 | history | edited | mucaua | CC BY-SA 4.0 |
edited title
|
| Aug 3, 2019 at 21:07 | comment | added | mucaua | Yes, I think it is meaningful to relate individual-level data to a population average, because it, if we can demonstrate a relationship, we could use the individual-level data, which is much, much cheaper and faster to produce, to estimate the population data for the many situations in the population data are missing. It is certainly not an ideal approach, but we either use this approach, or do not even find out if it could be feasible. We could relate individual level x data to create an average x data that is then related to the population level y data. | |
| Aug 3, 2019 at 16:00 | comment | added | Isabella Ghement | Why would you try to relate an individual level x value to a population average y value? Can you tell us more about that (e.g., is it even meaningful from a subject matter viewpoint beyond the explanation that you don't have access to the appropriate data)? What is stopping you from relating a population average x value to a population average y value? | |
| Aug 3, 2019 at 15:51 | history | edited | mucaua | CC BY-SA 4.0 |
added 6 characters in body
|
| Aug 3, 2019 at 15:37 | history | edited | mucaua | CC BY-SA 4.0 |
added 87 characters in body
|
| Aug 3, 2019 at 15:35 | review | First posts | |||
| Aug 3, 2019 at 17:43 | |||||
| Aug 3, 2019 at 15:30 | history | asked | mucaua | CC BY-SA 4.0 |