2

I'm pulling my hair out on this one, so hope somebody can help me out. I'm using SPSS 19 on a dataset structured like this (dummy data):

Time1   Time2   Time3   Group
-----   -----   -----   -----
0       2       4       0
0       0       5       1
1       2       7       0
1       3       4       0
etc...

I can easily manually or using SPSS calculate the means for the different follow-ups (Time1, Time2 and Time3) first for Group 0 and then for Group 1.

But then I create a clustered bar chart by choosing 'Clustered bar char', dragging Time1, Time2 and Time3 to the Y-axis, dragging my Group-variable to the X-axis. Which creates a chart like this:

enter image description here

Legend shows my three followups (Time1 etc), X-axis shows my two groups (0 or 1).

While it apparently looks fine and dandy, I can clearly see that the mean values are simply wrong compared to the mean I can calculate myself, or the ones reported when I compare mean using a T-test!

I mean the very last bar shown in the example should show the mean of Time3 for the last Group, right? That's what I'd figure, but the mean is not correct. :(

Is there something basic I just don't understand about the clustered bar charts? Any help or hints are very appreciated.

Nick Cox
  • 56,404
  • 8
  • 127
  • 185
Alex
  • 459
  • 3
    Not your question, but see e.g. http://biostat.mc.vanderbilt.edu/twiki/pub/Main/TatsukiRcode/Poster3.pdf for argument why these plots are poor choices. – Nick Cox Aug 06 '13 at 22:07
  • 1
    It is difficult to say without reproducing. So please give (a portion of) your data and the erroneous plot corresponding to it. – ttnphns Aug 06 '13 at 22:18

1 Answers1

3

It's a bit late but I just had the same problem and as I was reading your post realized what the issue is. It has to do with how missing data are dealt with, listwise or casewise. In Chart Builder go to Options and select Variable by Variable rather than listwise. The default seems to be listwise, so if there is a missing value in one of your columns, all the values for that row will be removed... I think. When you calculate the means manually I imagine you did it column by column and thus listwise deletion wouldn't be possible.

Jash
  • 31