I have four questions and I want to make it one variable. The one variable I wanted to make is work-family conflict. The coding I have 1= several times a week, 2= several times a month, 3= once or twice in three month, 4= Never happened and 0= Not applicable. I can here re code it and change it to greater the value, higher the conflict.
But to make one variable, should I add all the questions ? If I do so, I will get number ranging from something like 2 to 16, does it give any sense.In this case, should I exclude 'Not applicable' option, because I think, this option distorts the total value. And, If I have to exclude this, how can I do it in SPSS?
Suppose, I excluded this option and the range I got is 4 to 16, is it still meaningful ? Is it possible to recode as a low, moderate and high conflict? If it is not the right way, What else I can do?
In general, a 'not applicable' lowers the measurement type of your variable (hence from ordinal to nominal). Nevertheless, if you use a summation as you propose, it won't have any effect since you coded it as zero. Neverhteless, you can't have a real zero value then (for example when a respondent answers never, you'll have to start with coding that by 1, which is not the best option to go for). It's advisable to recode it to a 'missing value' (you can use any value not in your codeing scheme, but '9' is most used for limited scales (of course with no 9 as possible instance for your variable).
However, I would advice to change your coding scheme to start with. Have it represent the order of your answers (so from never to always -> 0 to 3 in case of 4 categories), this makes interpretation afterwards a lot more intuitive. If you code 'never happened' as 3, as you did, a 9 would indicate no conflict (opposed to what you state for 16). It would be more intuitive to have a higher score indicate 'more conflict'.
This being said, there are different ways to combine your variables. You will lose information using any of these posibilities. Assuming you know the variables are about the same concept, and all scales are in the same direction (always from 'never' to 'all the time' for example), you can indeed (1) just add them as long as you state the limitations of this approach (as you should always do). Imagine some forms of conflicts are more severe than others, (2) you could also attach a weight to the scores. The more severe conflicts will have more impact on the scale this way. A third option would be to (3) use a combination that allows you to bring the new variable back to its original components by using scores that always lead to a unique result for that specific combination of answers. The option you use depends on what you want to obtain from the result however.
There are other options though, but these seem to be the most logical based on your explanation.
You probably don't want to treat your "not applicable" as a 0. You should probably separate those questions from the numerical summation.
Think about the case where you are asking people about eating fast food. You have 4 questions. How often do you eat fast food? a) at home, b) at work, c) at school, d) while driving.
Person One answers answers 3 on each question for a sum of 12.
Person Two doesn't work, go to school, or drive. They answer one 5, and three NA, for a sum of 5.
But wouldn't you want the scale to reflect a higher score for Person Two?
If you divide each score by the number of valid, non-NA, scores, you get
Person One = average score of 3
Person Two = average score of 5
Also, the results are returned to the original scale of the questions.
* * *
If any person has too many NA's you may want to exclude them.
* * *
Obviously, the algorithm you use to compile scores can take into account questions with reserve scores or weighted questions.
There is a lot to learn about creating scales. There's a short discussion here in the Details section, that is a start.
https://personality-project.org/r/html/score.items.html
Dear Sushma
You can standrised your all questions and then you can develope a measure may be simple average or, if you know their relationship, you can develop a ratior.
Kind Regards
Rajiv
Yes, standarisation can be done by say (Value - min)/(Max - min) or (Max - value)/(max-min).
See whatever suits you.
Why not you use Netmeyer et. al 1996 Scale develop for Work family Conflict? that is widely used in HR and social sciences research. Scale is based on 7-point likert scale.
Netemeyer, R.G., Boles, J.S. and McMurrian, R. (1996), “Development and validation of work –family conflict and family – work conflict scales”, Journal of Applied Psychology, Vol. 81 No. 4, pp. 400-410.
Revisit your coding of the items, reverse them as you mentioned. But avoid using 0 in a scale. You can compute variable in SPSS or any other software, but the original logarithm behind computation of main variable is the average of the items.
In the context of SEM, you can impute the variable after CFA, but in SPSS taking average of all the items serve the purpose. You can recode, as 1 = never, 2 = rarely, 3 = sometime, 4 = often, 5 = always. By this when you take average, it would come around 3.
Any value high than 3 means conflict is high; any value lower than 3 would mean low conflict etc...
- Badges
- Science topic