Side-by-side bar chart for weighted data
Posted: Wed Jun 05, 2019 11:37 am
Hi all
Suppose I have a two-way table of counts, and the data are entered in "short" form (rather than one row per 'person', or "long" format); say
...where the columns are the counts, and the other two are the two qualitative variables and their two levels.
This is, of course, a very common way to enter such data, and SPSS handles it through a general Data> Weight cases option before doing any analysis.
Now, performing a chi-square test with data entered this way is easy enough in jamovi: Frequencies> Independent samples, and the "Counts" contain the first column of the above, indicating the number of times each of the combinations of the qualitative variables occur.
So far, so good... but I cannot generate a plot from within there (unlike, say, for a two-sample t-test). So I head to Exploration> Descriptives to try and produce a chart/graph. The two qualitative variables can be placed in "Variable" and "Split by" which almost does the correct thing... but the "Counts" (the weightings; the first column in the data above) are lost, so the plot is wrong. This would work presumably if the data file were set up with the one-row-per-'person' "long" format, rather than the combined/weighted "short" format as above.
Am I doing something wrong, or missing something? Or should this be a feature request?
Thanks for guidance; any advice appreciated.
P.
Suppose I have a two-way table of counts, and the data are entered in "short" form (rather than one row per 'person', or "long" format); say
Code: Select all
52,1,0
105,1,1
2,2,0
24,2,1
This is, of course, a very common way to enter such data, and SPSS handles it through a general Data> Weight cases option before doing any analysis.
Now, performing a chi-square test with data entered this way is easy enough in jamovi: Frequencies> Independent samples, and the "Counts" contain the first column of the above, indicating the number of times each of the combinations of the qualitative variables occur.
So far, so good... but I cannot generate a plot from within there (unlike, say, for a two-sample t-test). So I head to Exploration> Descriptives to try and produce a chart/graph. The two qualitative variables can be placed in "Variable" and "Split by" which almost does the correct thing... but the "Counts" (the weightings; the first column in the data above) are lost, so the plot is wrong. This would work presumably if the data file were set up with the one-row-per-'person' "long" format, rather than the combined/weighted "short" format as above.
Am I doing something wrong, or missing something? Or should this be a feature request?
Thanks for guidance; any advice appreciated.
P.