Data Prepping Tutorial: Cleaning Data
HTML-код
- Опубликовано: 20 мар 2020
- This tutorial covers the ten step process for cleaning data and preparing it for analysis for an undergraduate Marketing Research class.
Please note:
Due to time constraints and the undergraduate level of this workshop shortcuts were taken that are not recommended for more advance research. For example, I deleted the survey start, finish,completion dates and survey completion time. I used a quick and easy heuristic that if they finished the entire survey, that the data was good. It is important to know that this may create errors in the data. I recommend methodically reviewing any variables before deleting them. For example, variables like completion times that are too short (outliers or more than 3 standard deviations below the mean) may indicate a lack of involvement of the respondent and that the data is made up and not thoughtful.
Thank you for making this public. It was very helpful!
You are very welcome. Let me know if you have any questions and I will gladly help!
Great presentation. It's very important and helpful. Thanks
Honestly thank you so much this helped a lot
This was super helpful! I’m so grateful you went into in-depth details on certain parts!! I was wondering if you could help me I am kind of struggling on some parts of cleaning my data.
So I have an open ended question that I am having to code. There are 106 items participants can recall in their statement (68 correct items and 36 incorrect items). They’re split into 3 groups. How do I put this all into spss? It sounds so silly but I cannot wrap my head around it.
Thanks again for this video! :)
........Bravo Jennifer! 👍........
What an insightful Tutorial! 👍👍
☄Can you please introduce Visual/Statistical Reporting in Qualtrics via Stats iQ & Crosstabs? (Real examples would be great!).
Merci beaucoup!👍
Thankyou so much for the informative vdo dear
Hello Jennifer, i need some help please with cleaning my data, grateful for your assistance
What issues are you having? Also - where is your data coming from, for example Qualtrics, survey monkey, or a secondary source like CRM data?
@@JenniferTaylorPhD Its indeed my first time doing this i tried following your video but still nothing. I am using raw data from a survey.
@@corderotelesford1662 Were you able to download your data from the survey tool? Can you open it in SPSS? I am confused as to where you are in the data cleaning process, which makes it hard for me to help. Also, this is for a PhD, which the steps are a little different and more rigorous that what I present here in this video. But here are the steps, let me know where you are having issues:
1) Download data from survey tool to SPSS (make sure the file is downloaded in an SPSS format - this video will show you how: ruclips.net/video/ieAFNIx_1Eg/видео.html)
2) Open the SPSS file. Make a note of the number of variables and number of cases that you have. (you should have a document that lists each step you made with the data).
3) Assign unique ID's to each case.
4) Delete unnecessary variables that survey tool added - for your Phd You will want to keep Start Date, End Date, Duration_in_seconds as these variables will tell you if respondents were paying attention to the survey or if they just answered it without thinking (i.e, cases with very low duration_in_seconds may be removed. You would need to create a rule - such as anything above 2 standard deviations from the mean for this variable are removed). Again, make sure to keep a document as to what you did and the rules that you used.
5) Look at Progress variable. Determine what your rule is for removing cases. For my classes we use a 100% completion rule, but that is just to keep the math simple for them. You may decide that you want to keep surveys that were started but never finished. But if you are doing any kind of experimental work where a scenario is used to impact a dependent variable, then all surveys need to have completed the survey through the dependent variable. REmove any cases that do not follow your rule.
6) Produce frequency and histogram charts for ordinal, interval and ratio data (do not do this for nominal questions as it is meaningless). Look for skewed data and errors in the measurement items.
7) On your SPSS output, look at the statistics summary. Are there any variables that have "missing" cases. Do you expect these missing cases (ie., it is a check all that apply question) or is it an error?
8) In variable view, look at each survey question and review the labels and coding. Does the coding look right (i.e, very unsatisfied = 1; somewhat unsatisfied = 2; neutral = 3; somewhat satisfied - 4; very satisfied = 5). Or is the coding random, like 43. 44. 45. 46. 47? Make a note of any issues as this means you need to recode the data.
Once you have removed any unwanted variables and cases that don't meet your rules, and you have identified any variables that need to be recoded, then you are ready for the next video which is about recoding your data.
Recoding into Different Variables: ruclips.net/video/GnzAw_RzmdA/видео.html
Recoding into the Same Variables: ruclips.net/video/4cnS6JvRhIo/видео.html
Computing new variables: ruclips.net/video/FMAIWv6GgEc/видео.html
Let me know where you are stuck and I will gladly help.