0

Edit 1 -

Actually the "upcoming concept related to how employees work" is 4 day work week... i am trying to find perception of employees of 4 day work week...The employee has various attributes which included demographics and their work related attributes...and the perception part has attributes like whether if they like it or not, its advantages, disadvantages, etc. ..so i think it would be better if I combine all of my independent variables together and analyse their effect on the dependent variables together..not individually...as it would be more in line with my research question

I am doing a Research Project for my Master degree course.

One of my Research Questions in my Project is:

How do employees in Australia perceive the concept of <”an upcoming concept related to how employees work”> ?

I am trying to study the perception of this concept only w.r.t single country...

I need to take a survey in order to answer the above question. I have already prepared the questions but still not started the surveying part till now.

I will be doing this survey online..and plan to have close to 130-140 qualified responses..that is 130-140 responses who have stated in the survey that they are employees in Australia...hence they meet my criteria.. as per my Research Question...in order to get the required responses I will have to reach out to lots of people..which I will do from my end..

I was told by my professor that I need to do statistical analysis as well.

I have never taken any statistical class and have no experience or proper knowledge about how to go about statistical analysis.

After doing bit of research, I was able to find some information about concepts like dependent and independent variables, types of variables(categorical, continuous), etc. I also found that the stastical tests which I can do depend on the nature of my variables…that is most of the variables in my case are categorical…

After considering the variables, my survey questions and my research questions, I was able to come up with below-

Categories = No of options for the person to select from

Independent variables (Employee’s Demographic) –
Gender (2 Categories), Age Range (6 Categories), Relationship/martial status(3 Categories), Have Children ?(2 Categories),

Independent variables (Employee’s Work related attributes) –

Type of Employer(Govt, Non-Govt, etc.) – (3 Categories), No of hours worked per week (5 Categories), Work in shifts ? (3 Categories), Work in essential services ?(3 Categories), Work as a manager? (3 Categories),

Dependent variables(New Concept related attributes)- Like the concept ?( 3 Categories), Advantages(4 Categories), Disadvantages(5 Categories), 3 additional questions each with a 5 point likert scale.

I have still not started collecting the data till now...and need to be little bit clear about the analysis part..just in case I need to change anything in my survey...

In this case, can anyone tell me how can I systematically go about carrying out a statistical analysis for the above ? What all statistical tests can I do in my case ?

I would really appreciate your guidance as I am a complete newbie w.r.t statistics analysis. Any inputs or if any resources you can point me to…I would be really grateful..

Thanks, :-) S

Slim Jim
  • 11
  • 2
  • The question is unclear on whether you have already collected the data or not: "I have already prepared the questions but still not started the surveying part till now" versus "Considering that I have conducted the survey and have all of my responses..." How many countries are involved (you don't list that as an independent variable), how are you distributing your survey, and how many surveys do you intend to distribute/receive? Please provide that information by editing the question, as comments are easy to overlook and can get deleted. – EdM Jan 10 '22 at 20:16
  • Hello...sorry..my bad...i will update the question accordingly.. – Slim Jim Jan 10 '22 at 21:06
  • Edited my question.. – Slim Jim Jan 10 '22 at 21:23
  • At no point do you state your objectives. // Your dependent variable "upcoming concept related to how employees work" is inscrutably vague. Can you say more about it? // Then do you want to know if opinions about the 'concept' vary by age? by employer type? job type? // mainly one of these? each of these separately? combinations of these? – BruceET Jan 10 '22 at 22:52
  • @BruceET - Actually the "upcoming concept related to how employees work" is 4 day work week... i am trying to find perception of employees of 4 day work week...The employee has various attributes which included demographics and their work related attributes...and the perception part has attributes like whether if they like it or not, its advantages, disadvantages, etc. ..so i think it would be better if I combine all of my independent variables together and analyse their effect on the independent variables together..not individually...as it would be more in line with my research question... – Slim Jim Jan 10 '22 at 22:57
  • Do you want to know if opinions about the 'concept' vary by age? by employer type? job type? // mainly one of these? each of these separately? combinations of these? // With less than 200 subjects, I hope you can cut the number of indep variables and their levels to a minimum. – BruceET Jan 10 '22 at 22:59
  • @BruceET Edited my previous comment.. – Slim Jim Jan 10 '22 at 23:03
  • Actually the "upcoming concept related to how employees work" is 4 day work week... i am trying to find perception of employees of 4 day work week...The employee has various attributes which included demographics and their work related attributes...and the perception part has attributes like whether if they like it or not, its advantages, disadvantages, etc. ..so i think it would be better if I combine all of my independent variables together and analyse their effect on the dependent variables together..not individually...as it would be more in line with my research question – Slim Jim Jan 10 '22 at 23:04
  • 1
    As you've posted another very similar question on this site, I'd recommend that you close this question to avoid duplication. – EdM Jan 11 '22 at 15:27
  • I posted another Question which has all the details clearly to give context - https://stats.stackexchange.com/questions/560041/what-are-the-types-of-statistical-tests-analysis-which-i-can-do-here – Slim Jim Jan 11 '22 at 14:45
  • @EdM - Ok..will do..thanks – Slim Jim Jan 11 '22 at 16:15
  • @EdM - I can't find the option to close the question..can you let me know how I can do it ? – Slim Jim Jan 11 '22 at 16:28
  • I'll flag this so a moderator can do it. – EdM Jan 11 '22 at 16:41
  • ok..thanks...... – Slim Jim Jan 11 '22 at 16:55

1 Answers1

1

Although not stated explicitly, the statistical analysis is presumably to see how your independent variables (predictors) are associated with your dependent variables (outcomes).

Some of your categorical outcome data (e.g., the Likert items) are ordinal: the categories have a natural ordering from lowest to highest, even if the differences between levels aren't evenly spaced. You can take that ordering into account with ordinal logistic regression. Non-ordered multi-category outcomes can be handled with multinomial logistic regression. This UCLA web page has links to explanations of how to do both types of analysis with 5 different statistical software packages.

In terms of independent variables (predictors) in regressions of any type, it's possible to incorporate combinations of continuous, binary, multi-level unordered, and ordinal predictors. Other UCLA web pages accessible for the one I linked above describe how to do that.

There are some cautions.

For one, as you have multiple outcome variables, you might need to take multiple comparison into account. If you do enough tests of statistical significance, some results can appear "significant" even if they really aren't. The analysis must address that possibility.

For another, your sample size might be too small. To avoid overfitting*, you typically need to have about 10-20 cases per predictor variable. For categorical variables, each level beyond the first counts as a separate predictor in the model. On that basis I count 21 such predictors in your description. Some of those numbers come from your breaking down continuous predictors like age and hours worked into categories. That's a poor idea in general and increases your predictor count in this situation. But even if you cut down to only 15 predictors, your 130-140 responses would be barely adequate.

Finally, there's the big problem of whether your sample will be representative of the underlying population of interest. That might be the biggest problem of all in carrying out surveys.

*In overfitting, you might fit your data sample very well but get a result that doesn't generalize well to other samples from the same population.

EdM
  • 57,766
  • 7
  • 66
  • 187
  • Thank you for your detailed response. Since I do not have a background or any knowledge about Statistics...i am understanding some part of what you are trying to say...I will review the links which you have posted and write back...Actually this research question is a part of my Master Thesis(which is constrained by time) ... my Thesis has total 3 parts ... and this is one of them...I have designed the survey keeping in mind my research question...however i did not have an idea that the statistical analysis will be such a complicated part... 1/2 – Slim Jim Jan 10 '22 at 23:21
  • Now if i want to make it simple... and say that I want to group all of my independent variables together and analyse them against the group of the dependent variables(all together at same time)...can you tell me what is the analysis that I need to do..i will also address the limitations which you have pointed out...in this way my analysis will be in line with my research question... 2/2 – Slim Jim Jan 10 '22 at 23:25
  • @SlimJim multiple regression uses all the independent variables together. For the dependent variables, perhaps you can use your knowledge of the subject matter to combine all those individual response variables into a single measure. Otherwise, a separate ordinal/multinomial regression for each response is a straightforward way to go. I'd recommend talking with your thesis adviser about how best to proceed given the time constraint. I suspect that the adviser wants you to learn about how statistical analysis (and planning for it) is a perhaps complicated but necessary part of a project. – EdM Jan 11 '22 at 03:10
  • Thank you for your clarification. It seems that the way I am trying to analyse my question is way too complicated and maybe not even necessary...i have revisited my research question and reconsidering my objective..i think there is an alternative way in which I can my research question...i think it will be better if I make a new thread with all the details..i apologize for all the inconvenience caused...because of me not having any statistical background... – Slim Jim Jan 11 '22 at 13:14
  • I posted another question with all the details clearly mentioned -- https://stats.stackexchange.com/questions/560041/what-are-the-types-of-statistical-tests-analysis-which-i-can-do-here – Slim Jim Jan 11 '22 at 14:45