It is the mark of a truly intelligent person to be moved by statistics
—- George Bernard Shaw
Recently, I had found my self with some off time leaving me with some crazy hypothesis to validate and try out for myself.
One of my experience when soliciting opinions about my pic from my friends had an inherent bias in itself. My female friends had a favourable opinion compared to male ones. This led me to investigate if there is an inherent bias in the way the males reacted to the dressing fashion compared to women.
To check and validate for this hypothesis, I used a picture of mine and got it rated from Photofeeler for ratings from the community. The intent was to understand and capture if the hypothesis can be statistically validated.
Photofeeler is a means to get community votes to rate one’s pictures. You could either use earned karma to get votes or buy credits to exchange for votes.
Null Hypothesis: There is no difference in how different gender is perceiving picture attractive levels.
Alternate Hypothesis: The opposite(females) gender will rate the picture’s attractive levels higher.
- Male group sample size: 23
- Female group sample size: 20
- Parameter under investigation: Attractive
- Significance Test: Fisher exact test (small sample sizes, the underlying distribution is normal as well)
There is definitely some merit in how I’m being rated here. To further cement and keep the statistician in me, calm. I checked for one tail t-test (Fisher Exact).
The corresponding p-value is not significant enough. I guess, the sample size is too small. A repetition of the test with larger sample size would have enough power to probably reject the null hypothesis. As of now, there is no statistically significant inherent bias as I pre-assumed.
Still, looking at the numbers, it’s clear that the female populace was more forgiving of the attire compared to the male tribe. It’s a bit interesting to know the probable reasons why males’ rating was harsh.
This is just the start of the overall experiment with multiple iterations and factors to be accounted in next steps. The general idea was to test generic interest in something like this. You can comment out any other issues or concerns with the current setup.
In part 2 of the experiment, I carried out the same test with a different attire and the results this time were totally opposite.
The male population has voted in a comparatively favourable proportion compared to the women populace in this case. The numbers are too close, so no point running a statistical test, there is no real difference how the attire is being perceived.
This thought helps in clarifying a couple of aspects :
- The voting is largely dependent on the attire and not specific to a particular personality aka me in this case.
- Suiting up is a safe option though women were more welcoming of the hippy attire. This might act as a food for thought for the male populace.
- Your offbeat (Kurtha) attire might actually be fine and same gender opinions need to be taken with a pinch of salt.
I would still love to test this out for a different set of people and if anyone is interested, feel free to hit me up. We can see how this can be taken forward.
Another real-life data case study: Chegg
Another real-life data case study: Google Pay