In analysing qualitative data, we often identify categories and describe them in depth. My question, therefore, is: can the researcher make use of descriptive statistics to support arguments?
My other question is: when a researcher adopts a mixed method (irrespective of the design), is it adequate to use descriptive statistics to analyse the quantitative data?
Hi,
Descriptive statistics are useful in qualitative research to identify patterns and support findings and in mixed methods to analyse quantitative data.
Hope this helps.
The purpose of descriptive statistics is to generally describe the characteristics of a sample. It is necessary for the reader to be able to understand as to whom the study is actually about and what they are like.
In order to test a hypothesis, for example, the association between two or more variables, one has to opt for statistical tests.
Whether qualitative or quantitative analysis, the role of descriptive statistics is to provide the researchers summarises of the variables of interest so that decisions regarding the choice of model and data analysis are easy to make. For example, if the minimum mean of a dependent variable is 0 and the maximum mean is 1, the regression technique will be binary logistic regression. Also, if the dependent variable is ordered (1, 2, 3), the appropriate regression analysis is ordered logistic regression. Good luck
If I had an apple, I can describe it in as much depth as I want, but this alone does not justify to argue about apples in general. This particular apple may be special and not at all representative for what one should expect from apples in general.
If I had a sample of 50 apples, nothing actually changed. I have some additional information about the variance of features within my sample, but still this sample may be very special (e.g. all apples from some strange tree grown in a, say, unusually dry year) and not at all representative for what one should expect from apples in general.
All your arguments (based on data characteristics alone) always are limited to the statistical population the sample is taken from. It is the (assumed) representativeness of the sample, or, more specifically, of the sampling procedure.
If my sampling procedure was to take some apples from the tree in my garden that happened to be there last year, I can argue only about the apples of this particular tree in this particular year. If I defined a statistical population of trees (e.g. all trees in a sampled from an apple plantation and took a sample from all apples there at random (each apple of the plantation having the same probability of being sampled), then I can argue about apples from this plantation (in that year) even though I have data only about a more or less small sample of all these apples. Even a single apple (samples at random) would provide some statistical information! Each such randomly sampled apple will add some statistical information, so the total information obtained from a random sample increases with sample size. But if a take a single apple or large sample just because I saw it, I could reach it, and it looked so red and sweet, then this is not a random sample and provides no statistical information, no matter how large the sample is (unless the population is finite and the sample comprises a significant part of the whole population). So it's not in the apple - it's in the way the apple was sampled. I hope you get the point.
Joseph Baidoo , no! It does not especially depend on the sample size (if descriptive statistics support arguments). If depends on how the sample was taken. Only when the sampling procedure was random, then and only then does a sample provide information that allows one to argue (about the sampled population), and only then more data means more information.
Non-random sampling provides data that does not contain any (statistical) information about the population.* The information content regarding the population is zero. No matter how large the sample is, any sum of zeros will remain zero: no information.
I really wonder how you could get my previous answer so wrong.
*unless, as I said, the population is finite and the sample comprises actually the majority of the population. Its ultimately clear for a census, where the data from the entire population are known.
Yes, in analyzing qualitative data, researchers can use descriptive statistics to support arguments, Descriptive statistics in qualitative research are used to provide a summary or description of the data rather than to test hypotheses or make generalizations.
For Example: Frequency counts, Central Tendency, Spreads Etc.
Jose,
You should take into consideration that under QUALITATIVE research, where you construct the results, usually by situational review or discourse analysis, you DO NOT QUANTIFY.
Depending on the type of NON-probabilistic sampling used and leaving aside those single case studies. The number of participants is reduced, given that saturation of the category is sought. If you notice, only frequency and percentage is what you can do and talk about fashion at most. But to apply more, it is too much of an adventure.
In the Mixed case, if you want to do a quantitative process, it is clear that at least you are doing a cohort study, and therefore what you find statistically there, will only be from that particular sample.
to say that you can apply descriptive statistics and only use the mode makes no sense at all.
- Badges
- Science topic
- Similar topics
- Geoscience
- Atmospheric Sciences
- Atmosphere