Hi,
I am trying to run a multilevel model for binary responses and I am working through the worksheet provided by LEMMA. However I cannot seem to get a catterpillar plot. Everytime I reach this stage ths message pops up "Run time error 6 overflow" and I have no idea what it means. MLwiN then proceeds to shutdown and I have to start the process all over again.
Can anyone help me, is there something wrong with my data? Does my computer not have enough processing power?
I have 32265 observations and 119 variables. My levels are IDnumber and PSU.
Run time error 6 overflow
-
- Posts: 1384
- Joined: Mon Oct 19, 2009 10:34 am
Re: Run time error 6 overflow
Does the graph that you are trying to plot contain error bars? If so does switching to error lines fix the problem? MLwiN will currently give an error message similar to what you are seeing if you try to plot more than ~6000 error bars at once.
-
- Posts: 3
- Joined: Mon Mar 17, 2014 12:21 pm
Re: Run time error 6 overflow
I am just trying to to the simple caterpillar plot from the Module 7 (practical) worksheet for a null two-level model. At the level 2 I have 3,324 psus, where would I look to see how many error bars there are? I'm not sure what these are.
-
- Posts: 1384
- Joined: Mon Oct 19, 2009 10:34 am
Re: Run time error 6 overflow
In terms of the caterpillar plot the error bars are the "legs" of the caterpillar, and there should be one set for each of your level-2 units. I wouldn't have expected 3,324 to be too many however you can test this by manually creating the plot using error lines instead of bars. To do this firstly calculate the residuals as normal but instead of switching to the plots tab of the residuals window do the following (I am assuming that this is a variance-components mode similar to the ones at the beginning of LEMMA module 7):
Sort the residual columns (default c300-c308) by rank (default c305)
Graphs->Customised graph(s)
Under plot what?
set y to your residuals column (default c300)
set x to your ranks of residuals column (default c305)
Under error bars:
set y-errors+ to your SD(comparitive) of residual column (default c301)
set y-errors- to your SD(comparitive) of residual column (default c301)
set y error type to "lines"
Click "Apply"
If this doesn't work then I would be interested in seeing an example worksheet where this problem occurs so that I can check what else might be causing it.
Sort the residual columns (default c300-c308) by rank (default c305)
Graphs->Customised graph(s)
Under plot what?
set y to your residuals column (default c300)
set x to your ranks of residuals column (default c305)
Under error bars:
set y-errors+ to your SD(comparitive) of residual column (default c301)
set y-errors- to your SD(comparitive) of residual column (default c301)
set y error type to "lines"
Click "Apply"
If this doesn't work then I would be interested in seeing an example worksheet where this problem occurs so that I can check what else might be causing it.
-
- Posts: 3
- Joined: Mon Mar 17, 2014 12:21 pm
Re: Run time error 6 overflow
Hi,
I am so sorry it's still not working. I can just about get my head around statistics and then combining it with computers too...
Well anyway I followed your instructions and when I press apply it says that the "input columns have unequal lengths(sort)"
Followed by "group column wrong length dataset number 1" and the graph which is displayed is empty.
I can send you my worksheet if you think this would help?
Thank you so much for all your help so far.
I am so sorry it's still not working. I can just about get my head around statistics and then combining it with computers too...
Well anyway I followed your instructions and when I press apply it says that the "input columns have unequal lengths(sort)"
Followed by "group column wrong length dataset number 1" and the graph which is displayed is empty.
I can send you my worksheet if you think this would help?
Thank you so much for all your help so far.
-
- Posts: 1384
- Joined: Mon Oct 19, 2009 10:34 am
Re: Run time error 6 overflow
If you don't mind sending your worksheet I'll take a look. You can find my email address on http://www.bris.ac.uk/cmm/team/.