Hello,
I have two data sets. One collected by me(PRI_D) for one type of respondents (R1), other from a secondary source(SEC_D) for respondent types R2 and R3. I can further divide R1 based upon their demographies (individual factors) in R1a, R1b, R1c, R1d. I want to make comparison of respondent types R1 ( and R1a, R1b, R1c, R1d) with R2 and R3.
- There are 15 variables in total.
- Each variable is categorical (4-6 categories).
- SEC_D is available in form of %age for each category that can be converted to frequencies for each point manually. (Neither I have access to complete data-set nor I can get it on request-I did but wasn't given).
- PRI_D is available in csv file.
In order to make comparison of R1 (and R1a,b,c,d) with R2 and R3, currently I am using pivot tables in Excel to calculate frequencies for R1 (and R1a,b,c,d) and then using online Chi Squar calculator for comparing the frequencies for each category of 15 variables with R2 and R3, which is a tedious work and the margin of error is high as well.
Can someone recommend/suggest a quicker and less error-prone way out please?
I know how to use SPSS.
Thanking you in anticipation.
Br
Ali
I am afraid that there is no sound solution for your task, particularily when the source has some impact on the data, because source and variables are confounded and no analysis would be able to tell you if any effect/difference you might see is due to the source or ar "real" difference between the variables.
Thanks Jochen.
Actually in both data sets, sources (if I rightly understood you are referring to respondents?) are users. The variables are the actions they perform. My idea is to see if different user sets (students, experts, non-experts) perform similar or different actions. I used to same instrument for measuring the actions which was used in SEC_D study.
Do you think the source-variable affect will still be an issue?
If you want to compare R1 to R2, and if data collection (by PRI or SEC) can impact the responses, then it cannot be decided if any observed difference is due to the difference between R1/R2 or due to the difference between PRI/SEC.
A difference between R1 and R2 will only be interpretable when it is unreasonable to expect that PRI/SEC would have any influence on the respondents.
Thanks for the clarification.
Yes, mode of data collection may have influence on the respondents and this is one of the limitations of this comparative study. I will keep that in mind while making comparison and state as well.
It would then be really great to have at least one of the R's analyzed with both, PRI and SEC to get an idea about how bigh this influence might be.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS