Dear Prabin,

I think this is quite a question, because completeness of data may are interpreted as a marker of quality. Can you find out which data are missing? For example reasons for missing may be:

- Loss to Follow-up

- Complications during a hospital stay (in patients with complications the discharge letter is quite often not written)

- questionnaires, which are inconvenient for doctors or teh patient

It is possible to complete the data?

Best regards

Tanja

I'm totally agree with Tanja, you have to specify which kind of missing data you have, so you can decide the appropriate solution to solve this issue.

Hi Prabin

1. The Reliability consists of several measures:

Item alpha reliability and split-half reliability assess the internal consistency of the items in a questionnaire – that is, do the items tend to be measuring much the same thing? Which measure you want to develop?

2. All the above measures refers to the correlation between scores of any data variables, i.e. the correlation between two or more corresponding variables/observations.

3. Thus, you need to clean the data.

4. Please refer to:

Landau, S. and Everitt, B. S.(2004). A Handbook of Statistical analyses using SPSS.

2.Howitt, D. and Cramer, D.(2008). Introduction to SPSS.

Regards,

Zuhair

Dear Tanja, Ahmed and Zuhair,

Thanks for your response. In fact the data I am using is a set of secondary data related to sexual and reproductive health behavior. The variable of my interest, which is only one dichotomous categorical dependent variable has 22% missing data. It is not possible to collect the data again as it was collected some years back. When I did correlation the missing data are significant with p value less than 0.001 with all other independent variables. It is not possible to do another survey, and imputation can not be done as it is a national survey. I wanna know how can I justify to use this data? Can I say the missing data do not lie on any specif group of population but in all groups? Whether the missing data are MAR, MCAR or not missing at random?

Prabin

Dear Prabin,

first I wouls analyze the missing data - are these complete data sets or only data concerning follow up or special data (like questionnaires)? I would describe the missing data in the method part of your publication. In most cases, it is quite impossible to state, that only data from specific groups is missing. It is better to describe "limitations of the study" and to discuss it - otherwise you risk to be criticized by the reviewer.

Best regards

Tanja

Dear Tanja,

Thanks for you response. Its just a data related to specif question like have you ever had sex, the option was Yes and No. As my study is to explore the sexual experience, it is my dependent variable but nearly 22% of respondent did not answer which could be either due to respondent did not want to answer this or the interviewer missed to record. The information on all other independent variables are complete. Since it is national data and nation has already revealed the prevalence rate based on available (excluding missing data) so if try to impute, this may become controversial so I also want to just take a population those who have responded the sexual experience. I checked the missing with all other independent variables and found significant associations with all. I am wondering if I can go just analyzing the available data as a sample population. I will mention this methodology and limitation, however I would like to get some suggestion to justify this and probable question that may arise while presenting this.

Prabin

Dear Prabin,

I workes with a questionnaire in patients with heart failure. In this questionnaire, one question was concerning sexual function. This was the question with most missings. I looked analyzed who filled out this question and who not - the patients with advanced heart failure answered significant less frequent this question. I think it is the same like in your questionnaire - I would discuss it and I think that this will be an asset for your publication.

Best

Tanja