The current study Digital Marketing in Indian and its Impact on Consumer behavior in post and Pre-covid-19 in Agra a district of the state utttar pradesh India has to be taken up . How should I go about determining the sample size for the study please suggest the way forward.
Your help will be greatly appreciated.
There are sample calculators to assist with the calculation of the sample size. I generally prefer: http://www.raosoft.com/samplesize.html
Please remember all depends on the sampling method and your willingness to extrapolate the results on the population. Indeed sample size you may assess using sample calculators but there are two general rules to follow: minimum sample size for doing any statistical analysis (min. 30) and required number (min. 100) to try to draw any conclusions and talk about trends in marketing. Hope it helps
I think the sample size depends on the size of the target effect you care about reported in previous relative and similar literature if you can find it. Otherwise, you can determine the sample size according to a reasonable effect size (according to the effect size of the statistical method you want to use). I always use g*power to calculate the sample size and you can refer to the manual and application on https://www.psychologie.hhu.de/arbeitsgruppen/allgemeine-psychologie-und-arbeitspsychologie/gpower.html.
Rohit -
Be careful of online sample size calculators. They will make assumptions about the kind of data and methodology used which will often be completely inappropriate for your application.
You generally want to know what sample will give you a standard error that you want. Here it sounds like you are comparing values before and after an event, assuming this is a quantitative study, and you may want to look at confidence intervals around a difference in values.
One possibility: If you can plot continuous data, so that you have a straight line linear regression through the origin where the slope gives the ratio change from 'before' to 'after' values, you could see what sample size gives you an acceptable standard error for the slope, b.
In general for sample size requirements, a population standard deviation, or sometimes sigma from a regression (where sigma increases for larger predictions under heteroscedasticity of estimated residuals), is needed, and under a given methodology you see what sample size yields the desired standard error or better. For multiple attributes in your survey (more than one question) you need to check each one, or at least each essential one, for sample size needs and use the largest sample size (estimate) required for essential questions.
Methodology is important. For example, for stratified random sampling, you need a standard deviation from each stratum. Also, if you want to publish each group, they are no longer strata, but subpopulations which each need their own sample size requirement estimate. The overall sample size would then likely be larger.
Note that standard deviations are fixed numbers for which we need an estimate (or good guess), perhaps from a pilot study or other information. But standard errors are made smaller by larger sample sizes. That is important. It means we are assuming all kinds of bias are at least under control, and all we care about is reducing the standard error(s) by increasing the sample size.
Best wishes - Jim
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size