A crucial part to the answer to your question is whether missingness of data is random or not because your sample may no longer be representative of the population (if missingness is non-random) and may induce bias in your results. Assuming observations with missing values are randomly sampled, then the remaining observations are also randomly sampled, and your results are reliable.

Alternatively, you can explore the option of imputing missing data. Try estimation without and with imputation to see the sensitivity of your results.

A crucial part to the answer to your question is whether missingness of data is random or not because your sample may no longer be representative of the population (if missingness is non-random) and may induce bias in your results. Assuming observations with missing values are randomly sampled, then the remaining observations are also randomly sampled, and your results are reliable.

Alternatively, you can explore the option of imputing missing data. Try estimation without and with imputation to see the sensitivity of your results.

This is a known problem. The attached file may help you to identify the reason for the missing data, and whether it is possible to get around it.

For more advance information, you should contact "The Oxford Handbook of Panel Data," edited by Badi H. Baltagi, Oxford Press 2015. Chapter 5 of this book, p.152+, has a section on "Missing Observation in Panel Data" should be helpful.

MISSING data is considered a defect in ghat series, a simple solution is to remove that series, you have 391 of them, being left with 296 is no cause for concern. If any thing, it is a good sign of academic integrity and quality analysis to report that defective or missing data series had been removed. Any method of manipulation, no matter how tempting or theoretically sound, is nothing more than data manipulation.

ADEQUATE OBSERVATION is not at issue here, the missing data occurs in the cross section, not the number of observation. The number of observation is n = 17. Whether this is an adequate size of observation, do samole size calcukation for panel data.

Thanks for all valueble advices.I will follow your suggestions and will come up with what I achieved.

Another problem is is it appropriate to use eviews software for this kind of analysis.I choosed eviews since I am somewhat familier with it.

Software generally should not matter. And ideally, you want to use software you are familiar with. Good luck.

Consider the following articles;

1. https://www.academia.edu/1559425/Dealing_with_missing_values_in_microarray_data

2. https://www.academia.edu/17204529/Dealing_with_zeros_and_missing_values_in_compositional_data_sets_using_nonparametric_imputation

3. https://www.academia.edu/1240396/Dealing_with_missing_data

4. https://www.academia.edu/17922392/Dealing_with_missing_values_and_outliers_in_principal_component_analysis

5. https://www.academia.edu/1829576/Dealing_With_Missing_Survey_Data

Thank you sir

D.W. L.U. De Silva

One of the best software is Amelia.

Amelia II "multiply imputes" missing data in a single cross-section (such as a survey), from a time series (like variables collected for each year in a country), or from a time-series-cross-sectional data set (such as collected by years for each of several countries).

see: Article Amelia: A Program for Missing Data

D.W. L.U. De Silva ,

Eviews software generally omit the missing values and consider only complete observations in the estimation process. Still you can ipolate the missing values. And then compare the results with original panel results. But I would like to suggest that search for new independent variables and try to reduce the missing values or reduce the sample period.

Thank you. I will try both

Analysing missing data not only depends on the type of data you have but also on what you are going to do with the data. You could assign them case weights, or use the imputational techniques based on the kind of analysis you are going to use. Besides, you should also figure out the pattern of missing data if any in the data set, especially if the sample size is considerably big. You could also randomize your sample from the sample set and creating a sample form the sample. Lastly, if the missing data are only a few...just ignore them.

Best wishes!

I could solved the problem with missing values and panel structure with your suggestions.I faced another problem in panel regression.hausman test results revealed that fixed effect model is appropriate over random effect model.but unfortunately I am unable to perform a fixed effect model.im eviews it shows an error massage near singular matrix.then I checked for multicolinearity and remove two variables with high correlation with two of my most important variables and ran the regression again.but still it says near singular matrix.what could be the problem with my panle.need your help to overcome that issue as soon as possible since I am hurry to finish my research

Paul: I recommend you read this paper:

Missing Data & How to Deal: An overview of missing data

by Melissa Humphries, Population Research Center

Link: https://liberalarts.utexas.edu/prc/_files/cs/Missing-Data.pdf

Interesting question and follow up

The best way to deal with missing data or highly fluctuated trade data as dependent variable is to use Pseudo Poisson Maximum Likelihood (PPML) method because it has two benefits for trade data (i) it fits well in the semi-log model, so that countries with a small quantity and even zero value of exports some time would not be affected by the data; and (ii) it allows us to estimate the coefficients for time-invariant variables as developed in Silva and Tenreyro (2006).

panel analysis for gravity model approach of bilateral fish trade using E views. I have prepared excel sheet and analysis all the document

I think an analysis using unbalanced panel data is fine. By the way, it's far better than no data at all. You can have a broad-brush analysis by combining panel data analysis with cross-country analysis.

Regarding 'missing value' in data set, the critical thing is how to be sure that it is N/A (no data collected) or trade barriers (or other determinants) are so high that it blocks all traded transactions. This concern is particular relevant in trade paper. Unfortunately, it seems like it is difficult to know. But a close look at data is needed and, perhaps, the best thing we could do. Hope this help!