ASA Connect

There was a research on teacher perception of science in and out of the classroom and it was suggested that there could be a relationship between science teaching, nature of science and learners’ experiences of science in the classroom

But I have an issue with how the sample was selected for this research. Researchers used purposive sampling to select the sample. Data collection method were survey and case study. With the case study, the purposive is not bad but with the survey that is where I have the problem

The research covers two states. Even though it is a qualitative research, is it okay to use stratified sampling to select total sample before using purposive to extract the case study sample from the total sample? Thanks

------------------------------
Esinam Awuku
Statistics Officer
Ghana Education Service
------------------------------

Dear Mr. Awuku

For the case study, as you pointed out, probability sampling is not needed. For the survey it is a must. The results of the survey CANNOT BE GENERALIZED TO THE POPULATION, without some form of probability sampling to derive the sample. Now the type of sampling method depends on the structure of the population. Stratification is needed if there exists a population characteristic which affects the study variables. You mention "state" as a stratification variable. Is there a reason to believe that there are differences in perception across states? If yes, then stratified sampling based on state should be chosen as a sampling method. Are there other such variables? For example type of school, grade etc.? Then these variables should be combined to form the stratification variable.

Kind regards,

------------------------------
Yiannis Bassiakos
Associate Professor
University Of Athens, Economics Department

Original Message:
Sent: 02-13-2016 18:16
From: Esinam Awuku
Subject: Sampling method

There was a research on teacher perception of science in and out of the classroom and it was suggested that there could be a relationship between science teaching, nature of science and learners’ experiences of science in the classroom

But I have an issue with how the sample was selected for this research. Researchers used purposive sampling to select the sample. Data collection method were survey and case study. With the case study, the purposive is not bad but with the survey that is where I have the problem

The research covers two states. Even though it is a qualitative research, is it okay to use stratified sampling to select total sample before using purposive to extract the case study sample from the total sample? Thanks

------------------------------
Esinam Awuku
Statistics Officer
Ghana Education Service
------------------------------

Thanks for that.

The population was not homogeneous (it consist of experienced teachers on further studies, experienced teachers interested in their professional development but not on further studies and teachers who were neither interested their professional development nor on further studies). The target population was experienced science teachers teaching primary schools and these teachers came from two different states. The idea was to get experienced primary science teachers views on teaching the subject. In selecting the sample for the survey, the researchers picked from teachers who were on further studies for the survey. That is where I have the problem. In my opinion, stratified should be used to select for the survey at least. That I’m sure will create homogeneity or even representativeness and the final result can be inferred from.

------------------------------
Esinam Awuku
Statistics Officer
Ghana Education Service

Original Message:
Sent: 02-15-2016 03:16
From: Yiannis Bassiakos
Subject: Sampling method

Dear Mr. Awuku

For the case study, as you pointed out, probability sampling is not needed. For the survey it is a must. The results of the survey CANNOT BE GENERALIZED TO THE POPULATION, without some form of probability sampling to derive the sample. Now the type of sampling method depends on the structure of the population. Stratification is needed if there exists a population characteristic which affects the study variables. You mention "state" as a stratification variable. Is there a reason to believe that there are differences in perception across states? If yes, then stratified sampling based on state should be chosen as a sampling method. Are there other such variables? For example type of school, grade etc.? Then these variables should be combined to form the stratification variable.

Kind regards,

------------------------------
Yiannis Bassiakos
Associate Professor
University Of Athens, Economics Department

I don't see a problem since a stratified(assuming a description of how the sample was actually taken) is a very good method to use. In fact if they had stopped there and analyzed that data, they might have had a stronger result because they could have had confidence intervals that could be trusted.  Again, I'm assuming the stratified sample was taken as described in Cochran's Sampling Techniques. I am also assuming a purposive sample is a non-probabilty sample.

------------------------------
David Booth
Professor Emeritus
Kent State University

Original Message:
Sent: 02-13-2016 18:16
From: Esinam Awuku
Subject: Sampling method

There was a research on teacher perception of science in and out of the classroom and it was suggested that there could be a relationship between science teaching, nature of science and learners’ experiences of science in the classroom

But I have an issue with how the sample was selected for this research. Researchers used purposive sampling to select the sample. Data collection method were survey and case study. With the case study, the purposive is not bad but with the survey that is where I have the problem

The research covers two states. Even though it is a qualitative research, is it okay to use stratified sampling to select total sample before using purposive to extract the case study sample from the total sample? Thanks

------------------------------
Esinam Awuku
Statistics Officer
Ghana Education Service
------------------------------

Thanks for that. Actually stratified was not used only purposive (non-probability) was used in both instances.

------------------------------
Esinam Awuku
Statistics Officer
Ghana Education Service

Original Message:
Sent: 02-15-2016 19:35
From: David Booth
Subject: Sampling method

I don't see a problem since a stratified(assuming a description of how the sample was actually taken) is a very good method to use. In fact if they had stopped there and analyzed that data, they might have had a stronger result because they could have had confidence intervals that could be trusted.  Again, I'm assuming the stratified sample was taken as described in Cochran's Sampling Techniques. I am also assuming a purposive sample is a non-probabilty sample.

------------------------------
David Booth
Professor Emeritus
Kent State University

As a related aside, it should be noted that Jerzy Neyman's groundbreaking paper in 1934 was titled "On Two Different Aspects of the Representative Method: The Method of Stratified Sampling and the Method of Purposive Selection".

Relatedly, for those interested, Tommy Wright of the Census Bureau wrote an intriguing article titled "The Equivalence of Neyman Optimum Allocation for Sampling and Equal Proportions for Apportioning the U.S. House of Representatives", which appeared in the November 2012 issue of The American Statistician.

------------------------------
David Bernklau
(David Bee on Internet)

Original Message:
Sent: 02-16-2016 10:25
From: Esinam Awuku
Subject: Sampling method

Thanks for that. Actually stratified was not used only purposive (non-probability) was used in both instances.

------------------------------
Esinam Awuku
Statistics Officer
Ghana Education Service