- Posts: 11660
- Thank you received: 2742
Welcome to the LimeSurvey Community Forum
Ask the community, share ideas, and connect with other LimeSurvey users!
Dual Matrix with Exclusive Answer
- holch
- Offline
- LimeSurvey Community Team
I answer at the LimeSurvey forum in my spare time, I'm not a LimeSurvey GmbH employee.
No support via private message.
- Joffm
- Offline
- LimeSurvey Community Team
- Posts: 12944
- Thank you received: 3979
Is "numerical codes" suited only for "Dual Matrix" or also for "Array (F)"?
For answer options? In my opinion : Always.
The answer options of an array mostly are ordered like a scale.
So later you want to do some statistical tests.
With alphanumerical codes you are limited to tests for nominal scales, like CHI2
Joffm
Volunteers are not paid.
Not because they are worthless, but because they are priceless
- ymca
- Offline
- Elite Member
- Posts: 168
- Thank you received: 2
I'm so clueless on this subject, so I better change it to numbers if in the future I'll have a need for that.Joffm wrote:
Is "numerical codes" suited only for "Dual Matrix" or also for "Array (F)"?
For answer options? In my opinion : Always.
The answer options of an array mostly are ordered like a scale.
So later you want to do some statistical tests.
With alphanumerical codes you are limited to tests for nominal scales, like CHI2
Joffm
Thanks Joffm.
LS Version 4.2+
- ymca
- Offline
- Elite Member
- Posts: 168
- Thank you received: 2
I want to be certain so I'll know if to change or not.Joffm wrote:
Is "numerical codes" suited only for "Dual Matrix" or also for "Array (F)"?
For answer options? In my opinion : Always.
Joffm
You mean better to have it just for Array answers or every question type (in their answer)?
LS Version 4.2+
- Joffm
- Offline
- LimeSurvey Community Team
- Posts: 12944
- Thank you received: 3979
you can do whatever you want.
Codes are up to you. And there will be no impact on LimeSurvey.
But you might get in trouble if you want to do some statistic calculations later. Even the basics, like mean, median, std.deviation will fail, if you use alphanumerical codes.
Especially arrays usually have some kind of scale as answer options; you want to compare the different items, resp. subquestions. And here you need numeric codes, to be able to calculate a t-Test or a correlation or, or ...
Or can you tell me, what is the mean of "A1" and "A4"?
With alphanumeric codes, which are exported to SPSS as string variables (just do it and have a look), you may calculate "frequencies", "CHI2", tests that only need nominal level.
But to repeat: It's absolutely up to you.
If you use the "A1", "A2",... and later you find out, that you need numerical codes, you always may recode it in your statistic tool.
But it's unnecessary work.
Joffm
Volunteers are not paid.
Not because they are worthless, but because they are priceless
- ymca
- Offline
- Elite Member
- Posts: 168
- Thank you received: 2
My account > My personal settings
So now I need to change all the answers from all the questions to numbers.For answer options? In my opinion : Always.
Joffm
Also change the codes of the countries I have, to numbers?
LS Version 4.2+
- holch
- Offline
- LimeSurvey Community Team
- Posts: 11660
- Thank you received: 2742
1 Germany
2 France
3 Portugal
It
So if you have 2x Germany, 2x Portugal the mean value is 2, but of course the result is not "France", right? It's because this is a nominal scale.
So with all questions that you might want to calculate a mean value or something similar, you should use numbers. If you have a scale that is similar to a 1-3, 1-4, 1-5, 1-7, 1-10 scale, even if you use words instead of numbers, you used have numerical codes.
Definitely everything that is on a ordinal, interval or ratio scale. With nominal scales you can use numerical, but it is not necessary, you will probably not calculate with it, only count the frequency and that you can do with any code.
What you won't do is to have 5 male (1) and 10 female (2) and then calculate a mean value of 1,6667 here. It does not give you any real information. Just like with the countries.
www.mymarketresearchmethods.com/types-of...inal-interval-ratio/
I answer at the LimeSurvey forum in my spare time, I'm not a LimeSurvey GmbH employee.
No support via private message.
- miguel-reyes
- Topic Author
- Offline
- New Member
- Posts: 4
- Thank you received: 0
Thank you so much for your input. I have learnt a lot and was also able to remember some things I had forgotten.
I am now in the finishing touches of the survey and was thinking on doing some kind of validation so as to make sure that the user selects one of the songs with "Yes" in the "Select for Playlist" column.
How can I force this? I think most of my problems come from the lack of options available for the array dual scale question type.
I am uploading the whole survey so that this might shed some light for other people using the array dual scale question type.
Greetings and let's all stay healthy,
Miguel
- ymca
- Offline
- Elite Member
- Posts: 168
- Thank you received: 2
Good question.holch wrote: The countries don't need it. Because you won't calculate with those codes, will you?
I have no idea.
I imagine I'll want to make comparisons between year surveys (one survey from 2020 and another from 2021) and see the changes.
Also in which country there are more women that answered this and that...
And the different between women answers and men.
So for that I need numbers instead of letters?
LS Version 4.2+
- Joffm
- Offline
- LimeSurvey Community Team
- Posts: 12944
- Thank you received: 3979
sorry, but after ymca "destroyed" the original topic of this thread with his question about coding, I answer to his question first:
Didn't you read the link provided by holch?
There is a chart that shows which scale you need to calculate what.
Frequency, Mode,...: Nominal scale -> coded as you like
Median : at least ordinal scale -> numeric
Mean, Std.deviation,... : decimal scale -> numeric.
E.g. this:
Here you surely want to calculate some t-Tests (for independent and dependent groups) , Wilcoxon-Tests, Mann-Whitney-Tests, or ANOVA, ...And the different between women answers and men.
So, you really have to learn the basics about this, and these things are not related to LimeSurvey.
And to be honest: Sometimes I think you are just trolling.
Joffm
Volunteers are not paid.
Not because they are worthless, but because they are priceless
- Joffm
- Offline
- LimeSurvey Community Team
- Posts: 12944
- Thank you received: 3979
well, there is no question validation in array(dual scale). That's a pity.
Not tested yet, but as you show "group by group":
Could be possible to add a question at the end of the group where you do the whole validation.
Just a simple question, but impedes the respondent to proceed to the next group.
Validation could be something like
'(countifop("==","A6",that.A101)==1) AND countifop("==","A6",that.A102)==1) ...' (for the entire group)
Furthermore you have to include the check if the question was asked;
I think because of the presets you cannot check "is_empty".
It's an idea.
Joffm
I saw there is a javascript in these question.
Maybe now the count will be wrong, because you click two "Yes", one is hidden, but the count stays "2".
Did you check the answer table, which codes are saved?
Volunteers are not paid.
Not because they are worthless, but because they are priceless
- ymca
- Offline
- Elite Member
- Posts: 168
- Thank you received: 2
Sorry. I should have opened a new post even if this is the same subject?sorry, but after ymca "destroyed" the original topic of this thread with his question about coding
If I've been mistaken in the forum/post or something like that then please tell.
I did. I guess I didn't understand if to leave just numbers in the female/male and such, so I've asked it in a different way to explain myself better.Didn't you read the link provided by holch?
There is a chart that shows which scale you need to calculate what.
Frequency, Mode,...: Nominal scale -> coded as you like
Median : at least ordinal scale -> numeric
Mean, Std.deviation,... : decimal scale -> numeric.
And the different between women answers and men.
From this I understand I need to change to numbers.Here you surely want to calculate some t-Tests (for independent and dependent groups) , Wilcoxon-Tests, Mann-Whitney-Tests, or ANOVA, ...
Sure.So, you really have to learn the basics about this, and these things are not related to LimeSurvey.
That's not nice.And to be honest: Sometimes I think you are just trolling.
A troll is someone that is doing something deliberately negative and not to understand or learn something.
LS Version 4.2+