Has anyone experienced the error message when conducting multiple imputation of missing values in SPSS? In fact, I tried to conduct multiple imputation with my data and it appeared this message "
However, one of my variables is rated on a Likert-scale of 3 points and the other one are rated on a 4 point. Therefore, I treated them as ordinal variables, and therefore I think I shouldn't change the measurement level, or do you think I should change? What could be the other method of solving this problem, do you think that I should increase the number of parameters (which can take a long time) or do you think I should specify parameters. In fact, I tried to do this with one of my variables and the model run. However, I would like to hear your suggestions and advice. Can you please tell me your suggestions to solve this problem?
Many thanks in advance for your help and looking forward for your response
Filipa Calado
It is due to the scale problem. Change ordinal scale to interval and remove any interaction.
You can also check the missing value pattern in SPSS
Analyze>Multiple Imputation>Analyze Patterns
Hi Ankit,
Many thanks for your answer. However, do you think I should change to interval taking into account that it is a scale with 4 points? (and I also have another scale which has 3 points (likert scale)
. I also performed Multiple Imputation-Analyse Patterns and it appears this message "The Variable Summary table is not displayed because no variable has more than 10 % missing values" Therefore, what do you suggest me to do? Once again, many thanks for your help
You must check that the missing values are at random or is there any pattern (See the pdf link). If at random and you have Bigger data set (Let say more than 500 cases) then you may not need to replace missing values. 10% missing value is expected if you have community-based datasets. I usually avoid imputation of missing values unless it is very necessary.
You must convert it to scale.There are several control are provided in doing imputations for example can suggest max and min of imputed value etc.(See link)
https://www.youtube.com/watch?v=ytQedMywOjQ
http://www.sussex.ac.uk/its/pdfs/SPSS_Missing_Values_22.pdf
Hi Ankit,
Once again, many thanks for your help. I have read the link and the video you have sent to me. However, can you please tell me if you do the test Little MCAR test for you check if the missing values are random with qualitative variables? In addition, can I ask to you what is the method that you use when you don;t use imputation (that is when your data are missed at random), do you replace by the mean or by the most frequent value? many thanks again
Filipa
Multiple imputation is the best method to replace missing value till today. I think it takes account for both situation missing at random and systematic missing. While computing the sample sizes we usually make up for drop out or non responses for example if my sample size is 250 patients than i take 10% non response rate and tweak it to 280. after data collection i found 6% missing values and it is random, so I may not need to replace missing information. only if Missing values are so much that it reduce my actual sample size then we need to d something about missing information. Systematic missing are dangerous because it means you goof up the data collection (not always though as it depends on nature of the question too).I think SPSS had solution for checking the missing at random. MCAR test will be great to see that. You can go to Analyze>Multiple Imputation>Analyze Patterns in SPSS to check the missing value pattern.
Hi Ankit,
So, what do you do when you less than 5% of missing values and don't conduct multiple imputation. You perform listwise delection of all of these cases?
In that case, Missing values are not use in the analysis. Most of multivariate analysis requires to have non missing values for each of the cases on each variable (listwise deletion on included variables). In univariate or bivariate analysis cases with missing values on one or two variables are removed. Software will do it for you, you don't need to manually delete the information.
Hello there, just in case someone is interested in SPSS 23, it is now available in cloud with 30-day free trial and great price after. take a look at https://www.apponfly.com/en/ibm-spss-statistics-base?KAI
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS